Circus Chief is a tool for managing coding agent sessions from a browser. It's specifically optimized for small screens. It supports Claude Code, OpenAI Codex, and Google Gemini CLI agents.
Features
Agents can operate Circus Chief itself. Agents can spawn sessions, schedule sessions, interact with the Kanban board — anything you can do in the UI, an agent can also do.
Schedule work ahead of time.
Automatically reschedule when you hit usage limits.
Configurable, chainable prompt templates.
User-defined commands - agents can run them and see the results, and so can you in the UI. Handy for local CI or routine workflow steps.
Worktree-per-session isolation, or elect to work on a specific branch.
Shared canvas - this is a place for shared artifacts that don't belong in the code. Useful for iterating on planning documents.
Bring your own provider. Use subscription auth for Anthropic, OpenAI, or Google, or point sessions at third-party providers with Anthropic- or OpenAI-compatible endpoints.
I've built a stacked-PR tool for myself, it's just a simple wrapper around git commands. State stored locally in .git/config - https://github.com/alexghr/graphene
Lots of small dashboards/log investigations deployed to private Github Pages for $WORK. This has been a great way to share insights.
I'm currently working on a tool to control my tmux sessions from my phone. Specifically all the codex panes so I can remote control my running sessions from my phone while I'm away from my desk while also keeping them in sync with tmux for when I'm back at my computer. On this one I'm using Codex as a helper, it's not allowed to write any code so it's going along a lot slower :)
Both of the apps above was built using AI especially for the frontend. Both backend at some point of time I almost completely refactored/rewrote myself
Speaking of private apps:
- I made a small app which allows me to track business expenses and categorize them Built with AI and using AI.
But hands down the most useful thing I've made is HutchDB, which is a MCP service that you can call from any AI chat or Agent setup to store data for you.
Literally from your AI you just say "save that to Hutch" and then it figures out:
- The schema + fields
- Builds nice webviews (Kanban, Timeline, Grid, Calendar)
- Lets you share the output with people
So people use it for all kinds of things like time tracking ("every hour save a summary of my activities to Hutch"), for Agent to Human handoff ("Every day check social media for mentions of my company and save them to Hutch").
I use it for things like recording all of our marketing activities and then having my AI compare those to signups for rough attribution, etc.
https://concrete95.net/ - a musique concrète web app that's made to look like windows 95. Pulls random audio from freesound.org and loops a small section, you can also layer synth pads or melodic synth loops. I'm often able to get some really pleasant background ambience.
https://windso.me - a sample-based step sequencer that doesn't let you choose the sample that's loaded, kinda fun, still needs a lot of work!
- sandvault https://www.codeofhonor.com/projects/sandvault/ runs agents in a separate macOS user account, hardened with sandbox-exec. It also supports headless browsing and iOS Simulator from inside the sandbox for testing web and iOS apps.
- git-multi-hook https://www.codeofhonor.com/projects/git-multi-hook/ git only allows one script per hook event; this is a dispatcher that discovers and runs every script in a hooks dir, in parallel, for both global and repo hooks.
My blog is AI-coded: Zola static site, Sveltia CMS, Cloudflare Pages/Workers, with GitHub Actions handling builds and syndication. https://www.codeofhonor.com
Thanks to Push10K now I am on my 160th day of streak and 13,135 pushups (or my own version of equivalents) so far. It is the best thing I did after feeling like Glutton after dinner last Christmas.
I just found your blog yesterday reading up on the stories of how Warcraft and Starcraft were made!! Have been hacking on small games and a tool to build 3d environments for a while and get very inspired by hearing stories from back in the day. Thanks for making everything public. I really enjoy your writing.
> - git-multi-hook https://www.codeofhonor.com/projects/git-multi-hook/ git only allows one script per hook event; this is a dispatcher that discovers and runs every script in a hooks dir, in parallel, for both global and repo hooks.
Gemini has been indispensable for helping me move from Windows to Linux. I'm reasonably proficient, but moving to a brand-new OS brought so many random questions and weird edge cases that I never would have had the confidence, patience, or time to tackle it alone, even with pretty strong Google fu. It's been so nice to have instant access to answers for my specific problems, without judgment or having to wait on a reply.
At the same time, I moved from Chrome back to Firefox, and Gemini was great at finding equivalents for my most-used extensions -- and, when none existed, to write my own. It's also been really useful for customization/"ricing".
More recently, I got into Quod Libet as a primary music library manager, and both Gemini and Claude have been fantastic at helping me build custom plugins that make it do exactly what I want it to. Scripts to automagically download tracks with metadata and synced lyrics, a lyrics sidebar that highlights lines as you listen and lets you click to jump to a specific line, a bookmark button that lets you mark your favorite section of a track for easy browsing later. Next chance I get is something that enables lyrics search across the entire library (a feature I was already able to build for the Stremio desktop player -- it's so cool to be able to search for a line in a TV episode or movie and jump straight to it).
I similarly used it to finally facilitate a switch to Arch to a couple months ago. Not technically a tool, but it was very convenient to take the Arch wiki, and personalize it by saying "My hardware is X and my use case is Y, can you filter out the 90% of things from here that aren't relevant to me?"
instant access to answers for my specific problems, without judgment
For me that’s killer feature, even if we don’t achieve AGI at least we got good enough „something that will google it for me”.
It is great both ways as an expert in my niche I don’t have to waste time on reading through entry level fluff. As non expert in other fields I don’t get to be scolded for asking entry level questions so RTFM and LMGTFY will sink in history fortunately.
ChatGPT has helped me with innumerable little technical things and feels indispensable at this point.
I can't help thinking that this is a combination of Google sucking more and more and various problems of daily getting so hard need something like a script to solve them. (Recent challenge - "what affordable campgrounds are near the Pacific Create trail and open now").
I made a scraper that searches through all the news and finds anything 'war' related. It then summarizes it and give me that new in the Star Wars crawl with music.
I shipped a QA harness for Claude Code. Instead of clicking through flows by hand, it reads your code diffs, identifies the affected UI flows, and tests them in real browser
Plus after each run you get screen recordings with console logs, network requests, HARs, and Playwright traces so you can inspect exactly what the agent did :)
From a quick look at your profile, the majority of your submissions have been Show HNs. HN only allows some fraction of your submissions to be Show HNs (imagine if the front page was nothing but), so eventually they will just be auto-flagged.
I wrote a pen-plotter GUI and gcode sender in Rust. By hand. Like an animal.
I am the only user. Sometimes it's the process that matters, and exercising your brain is important too. I get that there is a lot of existential dread around AI taking our jerbs, and excluding humans from the process of creative work, but... you can still just write code, just for the personal satisfaction.
This is going to be one my next projects for experimenting with the Web Serial API. I got an old Ioline plotter that refuses to die. Any advise or tips for where to start with the SVG to Gcode conversion?
Many, really, but there are few I'm especially proud of:
- https://github.com/exlee/pikchr_pl - DiagramIDE (diagram amiga-style workbench where you can script Pikchr diagrams using TCL, Prolog, Pikchr or - recently mruby). Note: you need to navigate to actual crate for description. There are binaries built in case someone wants to try it.
- https://svg.axk.sh - semi manual SVG fitter so that I can easier vectorize AI-generated pseudovector images (who doesn't like 30kb SVG versus 1.8mb PNG?!)
- https://github.com/exlee/rik - this one makes me laugh every time I use it - it's an AI harness with text editor as an UI (i.e. it reacts to comment strings) - I gave it personality so it makes wacky comments but other than that it's very constrained agent (limited edition ;))
These are not vibe coded but AI made it much easier to slide through major friction points (e.g. for SVG fitter I really didn't want to reinvent fitting algorithms)
These are really cool - I've been meaning to return to my pikchr-in-wasm IDE experiment. I only ever built a PoC, and that was before AI coding. I should pick it back up and polish it a bit, but after seeing your DiagramIDE, I might just use that.
This whole thread is turning HN into my favorite app store. Good stuff.
A voice memo app, quite like the actual voice memo app from Apple. The thing is: now I can put my voice memo's on iCloud put Claude Code on it and make my transcripts into structured notes that my app then also displays.
So basically a way to just go on an hour long walk with myself, spit everything from the top of my dome stream of consciousness style, and then have Claude structure whatever I said.
It's nice to have something that structures my thoughts by just thinking out loud.
I vibecoded it (it's approaching 20K lines including tests). It works quite well but there are some bugs, so will have to do some actual engineering. But the UX is working quite well.
FWIW - In the Claude app (on android) when in transcript mode, it has a hard limit of 10 minutes. If you transcribe longer than that, it crashes and you lose everything you've said.
To make matters worse, they've recently gotten rid of the timer, so you have NFI how long you've been speaking for.
I use it for therapy-based stream-of-consciousnesses + venting and then have my project set up to understand the schema therapy work I've done with my psychologist and give me insights / draw threads between things from my past and now, and losing 10 whole minutes of talking and processing is SO FRUSTRATING!
It seems clunky for me. The dictating it self had weird bugs, like sometimes it decided to not work at all and the conversation mode is not very usefull for OP case if I am undestanding it correctly.
I am curious to check them out! I hadn’t heard of them. Marketing stuff takes time as I have noticed with aliceindataland.com [1]. Maybe I should do a show HN. I will think about it.
To be fair, vibecoding this memo app in Swift didn’t take too long. There were some tricks to it, using xcodegen helped a lot so that I don’t need to use the Xcode project.
It’s fun to see Swift code. I used to do some Objective-C back in the day.
[1] another thing I made. It’s a sequel to the Alice in Wonderland stories. It’s also a SQL course. I vibe engineered it, meaning I looked at the code and used AI-assisted development.
Except for the story though that’s almost all fully me. LLMs aren’t great storytellers. The same is true for the lesson scaffolding, that’s almost only me.
I've made a saas app to manage kids custody (mostly for France for now).
Turned out super helpful to keep track of everything in one place (shared expenses, ask / demand unplanned changes in custodies, kids activities (school events, birthday parties, sport competition and so on)
(Edit: forgotten in first edition) A cookbook to store the recipes my family likes to cook so I can eventually break up with Pinterest: https://github.com/vtbassmatt/Cookbook
I vibe-debug 10x more than I vibe-code, but here goes:
- a daemon to properly set the LEDs on my Lincstation N2 NAS (even early Claude was great at reverse-engineering): https://blog.majid.info/lincstation/
- most consequential, Thufir, a task manager to replace Cultured Code's Things as I am leaving massa Tim Apple's plantation: https://github.com/fazalmajid/Thufir Still very janky, but usable enough for my purposes
- most recently, a TRML dashboard for my Seeed Studio reTerminal E1002 smart e-ink display, and I also had to have Claude fix the firmware for full 6-color input: https://blog.majid.info/e-ink/
With Tailscale, you can basically point a domain to the FQDN of a machine you’re sharing with people and the domain will simply work for them (and only them)[1]. But for it to work without them having to know or specify the specific application port, you have to grant them access to 443 (and 80) in your Tailscale ACL for that specific host.
So yeah, now immich.familydomain.com works without family members having to remember the specific port. BUT, serveradmin.familydomain.com (another app on the same host) will ALSO be accessible to them (from a networking POV). We opened port 443/80 for that host after all.
I took a few hours with Claude back in January (?) and we wrote a tiny Go authorization gateway which basically consults both Tailscale’s public API and Tailscale’s `localapi` and returns the appropriate response to Caddy based on the requesting user’s actual allowed ports.
So now I can share different apps (subdomains) with different people without forfeiting access controls, all driven by Tailscale’s policy file.
(I hope I didn’t mess up the crux of it, pretty late here)
Edit: why not (something like) Authentik? Quoting from the draft:
I contemplated this for a bit but one thing I kept thinking about is:
_They're already logged in to Tailscale_
Why do I need to install and configure another full-blown app dedicated to doing something Tailscale already did? Why have the users go through two hops of authentication?”
Pangolin would have been a better direct fit than Authentik. It does come with an embedded idP nowadays but can also use Authentik. Another tool to keep an eye on is Netbird which is like a tailscale like mesh but has proxy support for whatever app you want to share with or without authentication.
Maybe…I really didn’t want to have to install another app just because I’m sharing though. My line of thinking was/is:
If tomorrow I decide not to share with anyone, I don’t want to have to reconfigure stuff. I simply edit the Tailscale policy file, and (maybe) spin down my server.
Before agentic coding went mainstream, one of my main use cases was creating sticker designs for concerts and music festivals. Creating the stickers and giving them away was a good way to meet new people.
I used OpenClaw to make a health and wellness coach agent that tracks calories and alcoholic drinks and logs it to a personal dashboard. I send it photos of my meals, and it will estimate the calories and log it. It will also help me make meal decisions and give me words of encouragement.
I used this HAM dashboard git repo to create a bespoke dashboard of different video and weather feeds from my area: https://github.com/VA3HDL/hamdashboard
I've spun up probably close to a dozen one-off or small websites for various little interests or projects. One of my favorites is a short domain file uploader that I can quickly host Markdown and HTML files to share with family, friends, and colleagues. It's using Caddy and running on a DO-VPS. I open sourced the code here: https://github.com/RobbyMcCullough/honeydrop
Made a multi-player Battlefield game for my scout group: up to 6 players set up ships like in the classic Battefield game. But you can try to hit any of the other up to 5 enemies and each team can see who hit them.
And instead of turn-based, you have to hike to/visit a physical location, fill in a code you found there to get bombs for the game. Or do a quiz and get the answers right.
I lost a camera recently and set up a Claude Cowork daily scheduled task to check if it finds my camera on a few marketplaces - Claude also goes through all the images and checks if it can see the serial number, which is engraved in the body. No luck so far, but we keep trying! It was incredibly easy to set up and uses my Chrome browser so it's logged into most platforms already.
I vibe coded myself a simple little Home Screen-installable webapp that tells me when my first meeting is the next day (I am in a lot of meetings, and they're constantly in flux). That way, I don't need to be logged into anything work-related on my phone, or even mentally engage in that world. I just want to know when I need to wake up. All it shows me is the time, and I can tap on it if I want to see the title. It adjusts the font and color according to how early the meeting is (earlier than 8a gets Nosifer).
Could I have done this myself? Of course. Would I have tho? Prob not.
This kind of simple, hyper-specific bespoke utility is the perfect thing for vibe coding IMO.
A SQLite based sweeper of all the scans, notes, PDFs and images I have on my filesystem, that stores their paths and allows searching their OCR’ed descriptions and text, as provided by Mistral OCR. I can ask things like “when does my car need maintenance” or “find me that picture my kid drew for Mother’s Day”. I use pi-based bash executable to launch a doc chat like that.
https://github.com/maxim/ringbinder
I built a browser extension that stops animated images such as GIFs by default.
I've always found it annoying that browsers autoplay animated images, and there still isn't a built-in way to control that behavior.
The extension shows the first frame and adds a play/stop button directly on the image.
What started as a personal utility ended up being published on the Firefox and Chrome extension stores. It's still a small project, but it solved a problem I had every day.
Nice, this means ebay will be usable for me again. A while back they started permitting videos instead of pictures for the thumbnails of certain products and I pretty much stopped browsing the site. As if online retails websites weren't bad enough already.
> I previously posted a Show HN about it, but it didn't get traction
FYI your linked submission is marked as dead. Not sure if that's a problem with your account or not. You should email the HN mods.
https://tajd.github.io/cofferdam/ - is an experimental compile time time checker that can be run for typescript/rust to be able to impose architecture, design principles that can be forgotten by agents. The goal is to be able to make it possible for an agent to be reminded about design conventions where it forgets, so it reduces the amount of context that needs to be used initially to define those sorts of conventions.
Hister is a full text indexer for websites and local files which automatically saves all the visited pages rendered by your browser. It provides a flexible web (and terminal) search interface with offline result previews & detailed query language to explore collected content or quickly fall back to traditional search engines.
It can provide a privacy-respecting search experience for serving "recall" type searches where users retrieve previously visited content, but falls short in "discovery" type searches (yet).
This was the first AI project I ended up working on as well, except I approached from building a meta-search first. I only added support for a local index recently (via SQLite FTS 5). But I haven't shared my project, whereas you have a truly fantastic webpage for yours. Plus going the extra distance with a terminal interface and MCP server too.
Much kudos. I hope more people discover how powerful even a local search index of previously visited content can be. And I hope more people can build large indexes as well, so we're not just relying on Brave & Mojeek & Marginalia (and EUSP) to rescue us for the fallback discovery searches.
Built a tool to help design cs/science inspired jewelry in CAD. I wrote a DSL to describe the jewelry, and had an LLM write the interpreter to generate a CAD file using cadquery (note to self: LLM suck at 3d reasoning). I would not have had the time to do this without AI.
Also used AI to design an online store (I'm not a front end dev). It's amazing to see my wife (non-technical background) tweak the web-site using claude code.
End result: an online store where we sell jewelry pieces that actually are lambda-diagrams (Tromp diagrams) that compute Graham's number, or of the Y-combinator (well, technically it's just a fixed point operator, one beta reduction away from the Y-combinator. But Y-combinator was not aesthetically pleasing from a jewelry point of view)
store: Built a tool to help design cs/science inspired jewelry in CAD. I wrote a DSL to describe the jewelry, and had an LLM write the interpreter to generate a CAD file using cadquery (note to self: LLM suck at 3d reasoning). I would not have had the time to do this without AI.
Also used AI to design an online store (I'm not a front end dev). It's amazing to see my wife (non-technical background) tweak the web-site using claude code.
End result: an online store where we sell jewelry pieces that actually are lambda-diagrams (Tromp diagrams) that compute Graham's number, or of the Y-combinator (well, technically it's just a fixed point operator, one beta reduction away from the Y-combinator. But Y-combinator was not aesthetically pleasing from a jewelry point of view)
I've found that Gemini Pro is surprisingly good at 3d reasoning. To back that claim up, I've had it create:
A WebGl program that takes input like X123 Y123 Z123 via WebSerial every 100 ms and builds an object out of the resulting path. Required some performance optimizations (just had to tell it what to do). Also asked it to make the corners nicer and it did. (To be fair, I'd already asked a lesser model and put some things in the prompt to nudge it the right way.)
Various OpenSCAD models. E.g., remote control holder with 5 slots, staggered heights, slight slant because it looks cool, and the slots all have different depths. One shot. It implemented the slant/tilt using a shearing matrix. 100 points.
What I often ended up doing is asking it to draw 3 labeled arrows X,Y,Z. So I could tell it to orient along the XY labeled arrows (which are in reality YZ, but whatever).
I needed something to help me stay off the computer (sites and applications) at certain times of the day with enforcement and in a way that's hard to remove. I had some ideas but was able to systematize all that into a proper program that I use daily. It's been very effective and it's much better (for me) than any of the commercial solutions that I've found for Linux. About 40% of the program was done using Aider (before I picked up Claude Code). The rest is using CC.
A self hosted web archiving tool with support for extendible processing pipelines (eg. extract article -> translate -> summarize -> generate tags, download video -> split audio track -> transcribe -> summarize), which led me to make a managed chromium browser with extensions and warc support for archiving, and a RSS feed synthesizer (take random article listing page that doesn't have RSS and generate a feed for it) so that I can plug it into my archiver. An active learning loop for a model to clean up articles by removing junk like native ads and sponsored blocks.
A tabbed terminal with project management features like launching the database, app server, and claude code in different tabs with one click, and split browser/terminal panes (eg. opening a browser automatically at the correct URL when the terminal reads http://localhost:4000/).
A modular MCP server with a MCP proxy and OAuth2 dcr so that I can easily add new random ideas for MCP servers in a few minutes with Claude and deploy them such that it's available to Claude by refreshing the tool list.
A small tool to render Claude conversations so that I can link to them from my obsidian vault with something like convo://claude-code/-home-jfim-projects-foo/<guide>
And overall just deploying docker containers for my self hosted setup
Most of it is on GitHub, in various states of readiness.
A simple one - a small MacOS status bar widget that shows me if my Ethernet adapter is the default connection or WiFi. It also shows the reported link speed. I have a _very_ flaky USB/Ethernet adapter so seeing "Eth 10" in the bar is enough to prompt me to unplug/replug and get back to "Eth 100" (yes, it isn't gigabit).
I've created lots of small things, but the most interesting is a firefox plugin that detects when some media is playing and send an event with the details to my home assistant. This way I can create some automations that automatically change the lights when I play/pause some media. It's really cool to have your lights automatically dim when you start watching a movie and then for some accent light to turn on when I pause the movie to go to the bathroom.
One of the "huh, didn't expect that to work" moment was getting GLM 5 to make me a user space driver for the Nintendo Switch Pro 2 Controller on Ubuntu.
When you plug it in, the device is recognised, but press any button and it attempts to start the pairing process. Then using evtest nothing is coming through.
That^ was pretty much my prompt too, and 10 minutes later I have a working driver with systemd unit so it works through restarts. Amazing stuff!
Besides, I made a lot of automation scripts (mostly using Ruby) that run on my raspberry pi to fetch/parse/crunch things and notify me on my Android phone through a self-hosted https://ntfy.sh server.
- A telegram bot that messages me in the morning and afternoon with a todo list essentially. It’s connected to Google Calendar and a crude memory database (SQLite). The kids wanted me to make it sound like the character Yarnaby from Silksong.
- Automated backups of steam saves for when my kids wanted to play the same game on my account and saves get out of sync with steam cloud. The kind of thing I wouldn’t usually bother with myself.
- automated script that reruns failed GitHub actions in repositories with flakey tests ‘cause why bother fixing them? It also auto catches up branches with the main branch for the repo.
- a YNAB extension for pi (agent harness) to help with entering purchases that need to be split across categories. This is also in the telegram bot so I can use voice-to-text to explain a purchase.
- I already had some python scripts to generate pdfs of Magic: The Gathering cards for printing proxies. I had an llm extend it to make some dividers with the set names and symbols on them. Makes organizing them a lot easier in the big card boxes.
Skills for creating good and repeatable benchmarking scripts.
A knowledge base for my research area, with tools for paper ingestion and search.
An md file to html presentation tool, there are several but this one helps me.
A review tool that splits a PR or branch intelligently into modules, and does per module reviews and global reviews for different aspects, and then summarizing that into a report. Can be used with multiple different harnesses. Written as a Python project, but build-time assembled into a single-file Python script with uv run --script shbang line.
I made a little tool where I can just drop HTML into a folder and it will deploy it either to my internal Caddy or publicly on Cloudflare based in folder. Can be a single html or a folder.
CF pages still required too much confusing clicking around on a webpage for me. This way I can just point any little report or app at a directory and done.
There’s others that are more server shaped and tightly coupled - a pipeline to pull in all my data like Garmin, Twitter bookmarks, messages into a Postgres DB. Kind of a personal data warehouse that i can use with apps/automations, like alerting me if my sleep schedule is drifting, and a custom web interface to my Garmin data
I got tired of all the quirks when opening CSV files in Excel, so I built a fast and lean viewer for CSV - at least this was what initially planned.
Later I find the technique I developed to instantly load arbitrary large CSV files can be generalized to work with any format, with an incremental parser combinator. This means the tool can read from not only files, but any stream-oriented sources like a pipe.
I did used AI in development but it didn't speed up the process very much, as I found a lot of time was spent on the deign of an intuitive and consistent UX. The project is still not very production ready, but in case anyone is interested: https://github.com/Verticalysis/Hitomi
It's been a dream to be able to work in the audio software space. I've been a musician, designer, coder all my life, and had a few moonlights with building audio plugins over the years. But now with AI at hand, I can use it to fill that missing technical requirement while still retaining my domain expertise (music/design).
I also built a kanban board with agent integration and context management, with a vs code extension to go with it (also helps with git worktrees too): https://www.agentkanban.io
There is AS Notes - an Obsidian / Logseq / Roam alternative for use a s a VS code extension (is designed for use behind corporate firewalls, git friendly): https://www.asnotes.io
AI has been a great 'exoskeleton' for me. I fortunately had some good infrastructure and solid application base templates from before AI 'got gud' and so building on these has been the best of both worlds - a solid base and improved speed of development.
My favourite football team was really at risk of relegation, then I created for me and my friends an MCMC bayesian simulator to estimate the relegation match by match.
It was an opportunity to get used in real life to some concepts (MCMC, Metropolis-Hastings, etc) that I always struggled to understand.
And my team got saved from relegation with an amounts of points very close to the number of points that my model forecasted
I also wrote an honeypot that emulate an Ollama instance: beside the attackers, it's funny how many people are looking for free inference. Somebody from Brazil try to use my honeypot to write to chapters of a book about traditional magic rituals.
My next step is to extract the data collected by this tool to extract IoC and malicious prompts and share them with the community.
In the same scope I wrote also an Ollama scanner: it fetch from Shodan the open Ollama instances, verify that they are reachable and check if they are real sending a dummy query.
Fiorentina, in Serie A.
After 11 matches we have only 4 points and we won the first match only in December: in the past nobody got saved with a so bad performances in the first matches.
I built https://beachcomber.sh after one day getting huge lag and asking claude to investigate found that one of the factors was thousands of resources purely for giving my prompt, tmux statusline, nvim statusline, claude statusline the same identical information.
I probably would not have bothered to allocate time to this pre ai, the juice wasnt really worth the squeeze. But I approached it with an initial amusing naivete about it being 'super simple'. As is almost always the case with software theres a reasonable amount of hidden complexity. But I have been using it as a sort of learning proving ground for how to work with agentic development. For example I got to a point where claude wouldnt implement properly and would argue with me about changes because it would read the current/old docs in the repo and get confused about reality. So right now I'm experimenting with 'canonical specs' that can only be changed modally with gates and a defined cascade from canon, to code, to docs in that order. Otherwise you end up in a weird thrash about the docs and the code disagree and which one will the agent decide to change for consistency?
Anyway, its been interesting and its v0.6 and at a point where Ive not hit a sharp edge dogfooding for a while and some beta testers would be valuable. Right now you have to manually wire it into your stack, once some others have kicked the tyres hard enough I will make some pr's to the popular tools to consider integrating it.
https://gitlab.com/grepular/foxcage - Runs Firefox inside podman to isolate it from the host. Has some interesting features that I wanted and nothing else gave me.
Currently working on a tool for sanitising email. Will be blogging it up at https://www.grepular.com/blog/ when it's ready for others to use. Does things like applying policies to html/svg/calendar/vcard parts to whitelist or blacklist tags/attributes/css/url schemas, clean URLs, fetch remote content at delivery time and attaching to the email to prevent tracking, pgp and smime auto encryption/decryption and a million other features.
A "Home Agent" setup, with customized special agents to manage various aspects of the house through home assistant, learning feedback from household users to try and tune everything at the right time.
Various MCPs for above.
A "remote claude code server", that gives project level overview and lets me run projects / develop on my home server rather than locally through my laptop. Supports ssh as well as a web UI (projects in a list, shell rendered using https://xtermjs.org/, with a tile overview when working on multiple projects to watch for turn ends.
Similar to above, I have a local version that auto launches a new project scoped podman container, passess through the work directory, installs CC/Codex/Grok into it and passess through the auth / config for each agent. Then dumps you into that shell with aliases to map each agent to that agent with a few special env flags to disable permission prompts, so claude = `claude --dangerously-bypass-permissions`.
An extensive MCP for Obsidian that gives agents access to use a lot of the more advanced Obsidian functionality, such as suggesting and installing plugins / configs / etc.
It's helped some of my daily productivity, but I still prefer to get my hands dirty with code most of the time rather than full prompt it.
Do you have any code/insights about this home agent you could share?
I'm thinking about creating something pretty similar, I want a digital housekeeper that keeps an eye on what is happening in my house and notifies me about dead/unreliable devices, fix broken automations, suggest new automations based on sensor data, etc...
I've setup the unofficial home assistant MCP already, but LLMs seems to struggle a bit to use it properly and I haven't looked into it yet to understand what is happening.
It's not just for myself, but I'm primarily creating it for myself - it's a browser for designers. I work in code but I often want a figma-type interface to explore different ideas without having to branch or litter my codebase with a bunch of demo components/files.
Normal browsers have built-in dev tools - this has built-in design tools. so I can visit my app, open up a surrounding canvas, pull fragments into the canvas, do some design-ish stuff, and merge it back into code. All in the same UI. It was cool enough that I'm going to release it, but for now it's very useful for myself.
A "plans for agents" interface because I'm tired of reading huge markdown plans to iterate on the real work.
Is supports infinite 2d space like miro, but has one thread per card, a cli for the agent to treat open notes.
It support voice prompting too.
It feels like you have parallel conversation with an agent, except it understands all the surface. And it's granular, surgical and precise.
Usage:
I'm writing a game design document of 70 pages with it, working surgically without having to worry about what page number the edit was.
It really solved my main bottleneck which is telling the AI what to do in a complete and comprehensive context.
Readme is trash but your agent will understand what to do
I was... not prepared for your standup to comic README. Besides just being a cool idea, you gave me some things to think about and several new rabbit holes to explore.
There’s another one I’m planning to start - subtitles for your conference calls.
It would generate netflix like expressive subtitles such as - “uncomfortable silence ensues”, “looks at a distance”, “playful music in the background” etc.
A year ago I was using claude and found it unexpectedly blowing-away already-settled code. I found this frustrating and just wanted to drop a "save point" where I could compare the before and after versions of certain files. So I built a little utility called "bup" that tarballs the whole code tree and gives it a unique name. It ignores prior backups and has options to ignore certain subdirectories, VCS directories, artifact directories, or secrets files. And you can add a single switch (-s) to create a shell script that automatically backs-up the current project. Yes I can use git for this, but this is honestly way easier, especially if I want to compare/edit different versions. No names. No files. No trees. Just "bup .". And it's 100% bash. https://github.com/maxk42/bup
I'm probably the only one who'll ever use it but I like it.
I made MooBlock - a browser extension for digital self-regulation. Basically, it adds a small timeout before you can access a distracting site. This timeout grows the more you visit these sites, and decays when you stay away.
And yes, there are lots of cows. The longer you stay on distracting sites, the more cows appear.
The idea is to make distracting sites less appealing, without using a black-and-white site blocker, which you can easily disable.
ASCII-Globe - https://github.com/jcubic/ascii-globe - JavaScript library that renders a rotating earth or any map. Can be used to add animation to your website.
I made a tool that creates sandboxes (docker, podman, orbstack, seatbelt, tart, containerd, kata, firecracker) and then sets up an agent (claude, codex, gemini, aider, opencode) inside it with max permissiveness (no prompts to call sed, etc).
It creates its own copy of your workdir for the agent to play in, and then you pull changes out ala git diffs or commits.
It's a MASSIVE time saver, and I use it as my daily driver.
The tool I'm most proud of is "Hex Flex" (https://seidleroni.github.io/Hex-Flex-Web/). It is a tool to view and compare the contents of Intel Hex files. Should be useful to other people who work in the firmware field.
Not exactly a tool, but I also made pelohard.com which ranks the most recent Peloton classes by difficulty. Updated twice daily.
I wrote a tui sql client to replace DataGrip (which is really slow). https://github.com/keithn/sql It's quite customized to what I wanted, I haven't really checked it works with other things.... only thing is, I don't really use it much anymore, I just get claude to do all my querying.
Most of the tools I write now are bridges to various SaaS products that have APIs but no CLIs.
I used copilot to make a basic MCP server for a large DB, so it could write better queries.
I also got it to write a mysql DB importer I could use instead of "mysql", since I often import large DBs locally. I wanted progress, estimated completion time, and I wanted it to prevent my machine from going to sleep during the import.
I've written my own programming language. IMO, "good architecture" exists outside of specific language choices - SOLID, various design patterns, etc.. I've always felt like I'm implementing the same high level design in any language I've worked with, it's just manifested and looked different, depending on the language tooling. The "good" opinions are baked into the _structure_ of the language itself, so the robots have no choice but to build well designed codebases.
* plugin for Logic Pro to A/B mix with reference tracks, with ai-based stem splitter (e.g. isolate vocals in ref track, and compare with your vocal track)
* plugin for Logic Pro to simulate how a mix will sound on my macbook and phone (I captured real impulse responses for that, sounds very close)
* an app for spaced repetition for guitar video lessons / their parts (no idea why platforms like truefire don’t have this feature)
* workout planning/tracking app
* an app to create impulse responses for acoustic guitar, to make it sound good live
I build a lot of data pipelines, and I've had to deal with too many inconsistent "source to target mapping specs" (usually Excel files) in integration and data projects in my life. They're too opaque for AI coding tools to get consistent results for generating implementations, suggesting tests, making test data etc. So I made "Satsuma"
(https://equalexperts.github.io/satsuma-lang/)
At first glance, it's just a nice version-controllable, parseable DSL. But I also made succinct prompts with the grammar that lets LLMs produce and reason about Satsuma, a language server, CLI tools for the AI tools to use to navigate specs token-efficiently, reason about lineage, pretty viz in vscode plugins/syntax highlighting, some agent skills etc. There are metadata conventions for succinctly representing a lot of quirky formats and capturing common analytics conventions (scd2/Kimball/datavault/etc c.)
Yes, yes, I may have gotten carried away.
But I'm finding it really useful as a specification tool in projects for reverse engineering mappings from code/workflows, generating new code (dbt, Spark etc.)
This definitely isn't something I would've had the bandwidth to push this far before AI!
I made a nice little CLI tool for testing the subjunctive in Italian. Claude code spun off a bunch of Claude API calls to generate example sentences with fill-in-the-blank spots for the correctly conjugated verb. Having an AI generate a prompt for and call out to other AIs was a bit surreal!
I've used AI for a number of Emacs-related utilities and configs. Just today I created a script to reproduce the particular combination of MSYS2 packages I use for my newer on Windows setup - the hard part being to get native comp working. Small, but it's the sort of rarely used convenience I wouldn't have written up in the past.
I tool which pulls all details from my Garmin and keeps a '5 day rolling average' series of stats -- my lifestyle means I can't keep up 'X steps a day', but I feel I can keep up 'average X steps a day for the last 5 days'. What's perfect is I can tweak it, I even added a special 'the weather is truely awful today, you can dip under average but you need to make it up later' option by getting it to automatically pull the weather forecast, or 'The weather is perfect today, really push'.
- An outreach planner composed in a single html file. Data can be exported to, and imported from, a JSON that can be passed to the LLM for enrichment. https://kashifaziz.me/outreach-planner/
- An analysis pipeline that takes data from D1 (storing user searches), Posthog and Clicky (analytics) to periodically generate user journey reports.
- halalcodecheck.com - a database of emulsifier codes and food ingredients. Building this with the goal to help consumer verify and find alterantives of food ingredients across US, EU / UK and Asia.
I made a small tool where I and my wife can send receipt photos, ask in text to create events or send poster of an event from Telegram and it directly sync it to my Google family calendar.
I've always wanted my own VW diagnostic tool suite, and between tooling that was released in public on GitHub (https://github.com/kartoffelpflanze/ODIS-project-explorer) and my own research from years ago, it always seemed straightforward but too tedious to execute on. Claude did a great job making something useful, https://github.com/bri3d/mcd-diag-rs , and now I don't have to find a Windows machine or remember a specific diagnostic cable to replace my brake pads.
I also build a ton of household glue stuff; I was never really passionate enough about the whole "homeserver" thing to spend the effort in going beyond basic video recording for my security system, but now I have all of my local-only home automation stuff wired together, mostly into HomeKit, and have been able to ditch a ton of cloud services.
I didn't build tools in the classic sense of something you build and run semi permanently, like baked into your setup or home server or whatnot; but I found myself building bespoke tools with most new projects at work. Get a new Jira ticket, figure out which components will be involved, often times the tool goes collecting logs and parsing them into a Web UI with buttons to toggle various features or params. And it tends to be different for every project. It's like the oldschool shell oneliners but more powerful and easier to write.
Countless little things to clean up data or improve tools I wrote long ago.
If you count HomeAssistant things (via MCP) then many many more little qol things, too.
Most recently, though, is a basic python CLI/Flask app that makes it point-click to manipulate the route table and dnsmasq settings on a raspberry pi. eth0 on the pi goes to LAN and put a USB ethernet attached to a switch and you now have a pretty powerful IoT/Untrusted device inspection environment.
One click to change the DHCP settings offered to the device, run a pcap, allow/deny the device access to certain services or NAT them elsewhere as needed. Straight forward DNS adjustments that can be applied per device and now that MITM-Proxy has a python API... it's pretty straight forward to also peep inside of SSL protected things.
Cloak windows from screen capture! Perfect for keeping things private while screen sharing over Teams, etc.! All other tools in the Store would trigger virus warnings when I tried to install them. Some were positioned to ask for money for basic features. My app is both free with more features and a more intuitive UI and set of keyboard shortcuts!
- usagi, for tracking Claude usage in the macOS taskbar. Lots of these around, but I wanted one that wasn't buggy and constantly adding features: https://github.com/duggan/usagi
- RockstarNinja, for sharing Claude plans and sessions. Since it's a bit of a data hog signups are limited, but me and my cofounder use it all the time: https://rockstar.ninja
- TweetEmbed, for browsing an offline Twitter archive and getting self-contained HTML copy/paste-able cards: https://duggan.github.io/tweetembed/
- bewitch, a terminal based metrics collector and data visualization system: https://bewitch.dev/
Each one of these has been a great way to really push on automating build processes, see what Claude can do for automatic documentation (screen captures, etc), and trying to give a distinct visual identity. In my next project, I'm trying to de-LLMify the prose it generates by using my own blog posts and aggressively pruning and integrating into the prompt.
The biggest change has just been unlearning "cost". Stuff I've subconsciously shied away from since intuition built up over a career has given me a sense of how long things take that just isn't true anymore. Still learning!
I made myself a silly ai-chungus it works as a series of containers that communicate over mqtt. I have an ollama shim for other services to talk to a model on my other machine, telegram shim that acts as a ui, a study component that will give me a random subject for me to study over the course of a week and I give a proper implementation (let’s say a ring buffer) and it reviews the code I wrote. It has some minor gen things using comfyui like a card pull system with cards, card text, rarity system, and special card effects like holo. And my favorite bit has been tying my todo system into it so I have a thermal receipt printer which will print a task I have to do for the day, prints a barcode, and when I finish the task I scan it with an iOS shortcut. It’s beautiful. What’s been the most fun was designing the mqtt topics in a way that makes everything else flow perfectly. Oh and there’s a tts system that uses kitten-tts which will produce audio for certain messages and an esp32 that gets those sound files and plays them on a speaker, or I can play the messages on telegram. Like if I do an /overview command on telegram i get an overview of my incomplete tasks and the ollama model helps prioritize. It’s my favorite use of ai junk at the moment
Sure so there’s a command in telegram you do /study and you can either tell it a subject or it’ll randomly generate one. I did /study ring buffers last time. So then opencode / GLM 5.1 or whatever write an entire site for me explaining the concepts, tradeoffs between things like infinite read write pointers, memory mirroring, how to tell if the writer looped the reader when the pointers are not monotonically increasing and instead get modulo’d, and bit masking with powers of 2 etc etc. then it wrote a ring buffer with various features in pseudo code so I can’t cheat. I also got papers to read from ACM and such. So I pickled it all over a week and came back with an Odin implementation, submitted to the site running on docker, and the ai reviewed it for me, telling me mistakes I made or areas where code isn’t really professional or I missed something. It’s genuinely super helpful. And the idea is whatever the AI tells me isn’t my only source of info so for example if the ai says infinitely increasing pointers will overflow but for example if I’m using Python, Python integers are basically bounded by memory and practically on most systems “infinite” so shut up AI I can do that or if it gives wrong info about something I’m getting published journals or websites as reading material that (hopefully lol) isn’t hallucinated. But really it’s the classic college style learn it then do a lab type of loop. Except it’s topics I’m genuinely interested in in the moment so the energy flows perfectly into it. My only wish would be to have it be cooperative working actively with others for the more accurate college experience bouncing ideas off other people seeing where we’re all confused or just me etc
I read "The 7 Habits of Highly Effective People" and wanted a webpage to set lifelong goals and tasks based on the four quadrants, so I developed a webpage.
https://life-compass-sand.vercel.app/
Always curious about which llms perform best in specific scenarios, so I built a local desktop app to benchmark and evaluate prompts and llms side by side.
I made myself a tool that connects to my cdj3000x and A9 mixer over network and gets all data live from them. So bpm, pitch, song metadata etc. The tool also connects to my recordbox library and runs custom ML algos to classify for pitch, stems, tonality Phrasing, energy etc. Long story short it basically shows me ideas of what will mix in well. It works bizarrely well for me.
if you have all of that info, could you then automate the mix transitions by sending MIDI commands back in? would be fun to play a playlist or radio but where the track transitions are automated, phrase-matched and non-trivial (perhaps lasting for a number of phrases)
And to test the hypothesis if the motivation actually works with "fast flashing images" I built this experiment: https://fitfatflix.fit/
When openClaw came out I build a game like map around it to orchestrate tasks: https://memefields.com/
And not published, I build a little recommender app of activities I can do only today or only this time of the year where I currently am, so I don't miss good things.
I built a route planner. I walk around the city where I live a lot and love exploring new streets, but at some point it became difficult to plan routes through streets I hadn’t walked down yet, so I created an app that lets you plan different types of routes. You can specify the percentage of overlap with streets you’ve already traveled, create a route from point A to point B, a circular route, or a one-way route. You can add streets to your favorites or exclude them.
The app runs only on my local computer. I use it all the time.
I’m working on Descartes[^0]. First to help diagnose what’s wrong with a machine. Later to help manage and monitor it by letting an agent build layers of tailor-made deterministic rules and statistical models, a bit like the description of the ship’s AI at the beginning of Absolution Gap. And serve as an ops "point of contact" for other agents for the machine / fleet of machines it’s in charge of.
Sometimes I daydream that the end goal demo would be to set one up in a VM and let a sort of ransomware lose in it, and see if the agent can identify what’s going on and react in time.
But for now I’m fighting Apple notarization to enable local notifications on macOS.
I wrote a TUI to search for words/phrases in a crossword dictionary. I’ve never written a reactive TUI before, and never written Go either. I was very happy with the assistance rendered by ChatGPT.
Of course, it was great to have exactly the features I want, and I enjoyed learning some new things.
I have made a TUI tool for organizing and authoring sermons written by myself. It also has a third mode for reading one or two bible versions in parallel. I love the interface and keep adding features to make it easy to navigate in and something that actually helps me to organize and write sermons. Some features are to have notes at specific places in the bible, summaries, notes, exports to html/pdf, metadata for each sermon and autofocus on widgets when changing between the three modes. Happy to work in the terminal this way. :-)
I built a mac native PDF reader called Xeil with toggleable darkmode that preserves images (inspired by veil), vim keybinding and fuzzy search. Before this, I was using Skim with inverted color to simulate dark mode.
Instant windows switcher with custom shortcuts and instant "opt+tab" and trackpad switching. Simple does exactly what I need it to do (just bypasses the slow window switching that is annoying), with no additional features or bloat.
I've also done a TUI that combines my messages from WhatsApp + messenger + discord which is pretty handy at work.
The front camera apps always show the mirror preview. Most of us hate photos others take of us, but love our selfies. It's because we groom ourselves and find our perfect angles using the mirror preview. So it's jarring when we see the photos of us taken by others. I always wanted a camera app that showed the non-mirror preview. Surprisingly most camera apps don't have this option. So I created this app to scratch my own itch:)
Note: Photos and videos are only stored in the browser. No data is sent to any servers. You can also install this as an app from your browser since it's a progressive web app.
Swift/iOS: An app that helps me track how much time I spend in the five heart rate zones, so that I can better focus my workouts. I'm working on version 2 right now, which will take advantage of some new features in iOS 27 and has an all new UI. https://www.zone2.app
I made a little script that takes all images in a folder and formats them for my digital photo frame. It knows which have a converted file in the dest and skips them.
I built a little Gmail cleaner because my inbox had a bunch of "newsletters I unsubscribed from in my head but never actually did." Headers only, never message bodies, because privacy should still matter, I think. It worked very well and the amount of spam I get dropped dramatically, which was the point.
Everything was built with AI as part of a 12-in-12 challenge I'm doing, where I'm building 12 products in 12 months. https://twelve.zamith.pt/
is:unread -is:starred <-- go through your inbox and star what you want to keep, this filter will help you delete everything else unread. Add something like older_than:1y to also prune your Gmail from time to time.
I was running a DnD online game and didn't feel that all the battle maps out there were easy enough to use. I am running using a screen share on discord so all the extra features of Roll 20 for example just got in the way. So an hour with Gemini made the rather bare bones map that I have here.
I've made a number of ceramic molds for slumping fused glass into bowls. As well as wooden templates for ceramic mugs. I've devised a few carrying tools to move glass frit paintings from my studio down to my barn where the kilns sit without spilling the glass.
Or were you only asking about digital tools? I haven't really made any of those. Making physical tools feels much more satisfying these days.
> Or were you only asking about digital tools? I haven't really made any of those. Making physical tools feels much more satisfying these days.
I made that realization last year and since then it's just been random project after random project each one requiring me to discover a new tool or method to do some aspect of the project "properly". I'll never be a plumber or electrician professionally but it's so rewarding to start from zero and learn something new that is tangible in the real world. That's the one AI use case I've walked away from feeling like I actually learned something.
You can do interesting things with microwave kilns these days, I wonder if they get hot enough for ceramics? They can melt copper, I believe, so they’d be in the ball park.
The overlap between ceramicists and technologists is never zero. Part of my initiative to slowly replace every part of my house and home with things I've made... 1% progress is good, right?
I'm so glad i'm not the only one....well I have dreams and visions of a plan but the most i've done is a half baked Patreon and Substack scraper that only kinda works to capture my sunk cost of subscriptions I never used, a movie theater listing app that allows me to find classic movies that may get buried from the mass advertised slop, a custom sewn jacket that contains pockets for homecooked popcorn and locally grown fruit and well a 3-D printed sauce cup holder for all those sauce cups that I get from fast food restaruants.
Im slowly trying to extricate myself by cooking more from home only from local farmers and what I can grow from home (so far only one cucumber). After all, can you really build everything else if your own body molecules are being replaced by low quality things made by others?
I'll get around to 100% at some point before I die or I wont care anymore since i'll be dead: one of those outcomes is inevitable.
I'm ahead of you by a ways but your instincts are not wrong :) we have timber and plans to replace the flooring with logged and milled timber, sourcing clay from the property, making tomato sauce of the gods from home grown roma tomatoes. it's a lovely way to spend a half century I think.
Last year I became the volunteer coordinator for my kid's school PTA. They had been using Sign Up Genius for years. I spent a full year fighting a paid tool, gave up, and re-wrote it.
Now https://voluntold.fyi exists, and I never need to remember to manually move my single $100/year "ad free signup" off an event that has already happened to one that is coming next week.
Looks really cool, its nice to have an alternative to the big players.
The only criticism is that, in the UK at least, "slash" can mean urinate, and that's all I can think of when I read through the website. Probably a me issue though!
Spent a lot of time initially with asking it to generate me radically different approaches. Then I cherry pick what I like, iterate over and over again.
It's still something I ask of it regularly.
Thank you for the kind words. I've always struggled with UI as I have a backend engineer background, now more on the CTO sides of things. I've always lacked the creativity IMO, and this gave me a very good opportunity to practice and try to build something ""unique"".
I had a portal/app with messages exam results, calenders for
.) kindergarten for my youngest child (who will now go to elementary school)
.) elementary school for my middle child
.) daycare for my middle child
.) high school for eldest child
All spamming with messages and news - so I made python scripts that use the apps API to get the messages (exams and changes etc) and puts it into our family telegram group chat if relevant
I just can't sit here every morning 20 minutes on my phone logging in checking everything just to find out the food plan has changed ...
Quite a few during the early days. Recreating some fun ‘game’, popular during the Flash days. A work breaker/silencer, etc.
Then, I got bored because they seem to be bad beyond certain complexities. But around Christmas last year, things improved a lot, and I’m getting confident building real ones. In the mid of these, I also got an offer to work with the [pre]sales/GTM team of a large company. So, I have been building working prototypes of bits and pieces of boring enterprise business around the world and have been racking in, if not billions, but very close to it in sales pipeline in about two months. (I did that 1B token in a week thing.) I’m sure the business and sales team will be able to convert 20% to 30% of that in the coming months. I also pitch and presented my work directly to customers and they are coming for more. So, it has been fun.
Lastly, I really wanted to scratch an itch I had for a while—build a Static Site Builder. But I ended up buildig it to be a documentation tool for TypeScript and JavaScript. It can be used as an auto-detect and let it build alongside front-end projects, make it a hybrid so you can add human-written documentation too or just build a Jekyll-esque static site.
https://github.com/LukaWe/LocalExifGeoMap
LocalExifGeoMap - privacy-focused, browser-based tool to visualize and analyze GPS data from a batch of photos. generate interactive maps, heatmaps, elevation profiles, and trip statistics. etc.
https://hrvsync.nglx.io/
Simple iOS app to sync Garmin HRV to Apple health. It's nicer to see the trends in Apple Health - I know that the measuring is different but that doesn't really matter if you not don't mix it.
- A HTML prototyping skill and a simple upload web app to share it privately with my team
- A Chrome extension to make design comments and ASCII wireframes right on the page you're working on and feed it to your AI coding tool as a prompt https://getdesignjam.com/
- An iRacing overlay to compare the telemetry of a reference lap against your live data https://flylapsim.com/
It's been a ton of fun to be able to dabble in software thanks to AI. Especially for small personal tools like these it's so powerful
I've got really comfy `just` scripts for generating Clang "intermediaries" in my CMake project. I can generate `.ii` files which get formatted and edited in a manner making them directly recompilable, along with `.ll`, `.bc`, and `.s` files. All the above are per-translation unit or post-LTO and I can constrain the output to specific functions or files, and the LLVM bitcode can take optimization passes or optimization levels to very easily introspect how my work in this codebase optimizes.
I've also got a clang-repl wrapper for this codebase that is very easy to use and makes interactive programming much easier for me.
I made my own agent harness inspired by OpenClaw, but with internal group messaging between agents and memory sharing confined in agent groups. Plus Telegram integration with multiple agents in one channel using topics: https://bazilion.com/
I made mobile friendly agentic driven IDE (as a PWA), where I can start "agent" processes on my different laptops (company, personal) or on the dedicated server and manage them from any browser / mobile device enywhere. Now I spent most of my time do programming in my head walking somewhere and talk to agents / review code directly from mobile. I also added support of n.eko like remote browsing streaming for emulating desktop browser while I'm on mobile and simplified deployment tool to manage kubernetes / docker deployments.
1. Comprehensive tool for auto expense logging and management, expense trend analysis, budget allocation, expense divider during group spend, report generator etc
2. DashCam app for simultaneously detecting threat and recording video.
3. Stock market portfolio management. A comprehensive tool which takes stock market investments as input, analysis the investments, provides a complete analysis, trends, suggestions etc
4. Fitness app. Records calories gained/burned,physical activities, all health parameters like Bp, Spo2, Blood sugar, Heartrate, Weight, Allergies and other synptoms. Analysis the health trends and provide suggestions regarding food, exercise and other health related anamolies.
5. AI learning app series ((13 apps to help learning AI from scratch)
6. Private chat app using Bluetooth communication
7.My own versions of Doc scanner, phone tracker apps
8. Health app which analyse tongue, eyes, and any symptoms to suggest the possible health issue and the remedy
9. JEE preparation CBT app
10. Electronic circuit builder and simulator app. (All basic circuits using AC/DC power, resistor, capacitor, transistor, diode, led, zener diode, switch etc can be created and run, it supports multimeter and CRO tools to measure different current/voltage and watch the waveforms)
Lot other things are in pipeline and will post once i complete them.
I then used it to build other tools -- a personal time tracker and a Wesnoth-like game (both not yet published). Basically, I maintain a Markdown file with a queue of tasks, and I run a Jaiph workflow in a loop that automatically picks up tasks one by one and develops them with no human intervention.
Too many to count. Most recently, an Alfred workflow for opening my IntelliJ projects either in an an IDE or terminal that also comes with an integrated build task runner, so I can quickly discover and run build tasks even when I don’t have a project opened anywhere.
After dropping the keyboard for the TV one too many times, the touchpad stopped working.
I looked for a decent remote keyboard app to use on the tablet, and found nothing I liked.
I ended up asking an AI to make something that served a webpage that connected back to itself via websockets. It provided a keyboard and touchpad on the webpage and forwarded events to uinput.
It works well enough on tablet and phone that I haven't got around to replacing the original keyboard.
Amongst lots of little tools, fed up of scribbling down my son’s football (soccer) scores in the Notes app, I cobbled together a little web app instead:
a TUI that runs all my daily Linux updates from one place.
Updating my machine used to be a dozen scattered commands: apt, fwupd, and
a pile of dev tools I want kept current (node/npm, go, claude, opencode,
plus binaries like kitty and lazydocker) that each update their own weird
way. Now it's one Bubble Tea dashboard that checks each tool's installed
version against upstream and only downloads what's actually behind, so I
can see at a glance what's current and what's not. Adding a new tool is
easy enough that I just toss new ones in as I start using them.
It also has a cleanup mode that hunts down all the pesky cache files that
quietly eat your disk and reclaims the space in one pass, which has saved
me from the "laptop full again" scramble more than once.
Currently, testing it privately and tweaking to make it awesome.
Another tool I made for myself is automated video editor (takes folder of raw photos images, and generates edited video that can be shared online). Used it among other things to edit all my GoPro raw footage laying around (hundreds of files).
I was always experimenting with different agents and models, and I wanted to track how much I use eg codex vs opencode vs hermes, and which models I use.
So I made, kind of last.fm/waka-code for agents where I track (anonymous) usage per project
Data is fully anonymous, all I collect is (agent, model, project name [can be mapped to something else to hide the true name]) and everything is opt-in by default (per project)
- a proprietary camera library that communicates with industrial-grade digital cameras.
- a full on self-hosted lab management system (journal articles, obsidian-like notes, lab notebook, kanban, embedded dna editor, LIMS), obviously this builds on all of the above items
An inhouse orchestration tool to run coding agents. It's so useful. I used to use tmux to run my coding agents and have little scripts to help me manage the workflow so this tool lets me encode my workflows and preferences. eg. I prefer to run sessions for the same project serially, working on main, rather than in parallel like how Conductor does with the help of worktrees.
I've been making a set of local- and files-first iOS apps for myself - Photo Gallery, Music and Contacts.
I use Linux on my computer and an iPhone and wasn't happy having to use a cloud to sync my data between my devices. Syncthing/Synctrain + my apps allow me to keep everything on my private network, without ever touching a server or having to self-host something like Immich.
- A few Tmux plugins to automate workflow tasks (jump to pane, resize panes)
- lots of TUIs for displaying relevant information from services (github, jira, obsidian), that are tiled on a "home" tmux window, gives me a live dashboard of work that I can refer to over the day.
- TUI games, started with a simple dope wars clone … for when I need a quick break.
Lookup or modify selected text using AI (chrome extension). I just select any text and click the tiny popup button "what's this" and get an answer right there on the page. Made it mainly to explain terms and abbreviations I come across on HN often. Can also ask any other question about selected text. Can even modify the selected text the same way. [1]
OneNote to markdown/obsidian canvas converter. It did that using interop api to read the actual XML of the onenote files.
Work time tracker as 1px line on edge of monitor. Shows thin line at the edge of the display which fills up based on what i am doing.
Plaintext bookmark chrome extension that save links to local markdown file, Dynalist, Workflowy, Github Gist and import export between them. Was originally for Dynalist when AI couldn't do much 2-3 years ago. Recently added these other end points. [2]
A heart rate monitor with finger on camera. It's bit crappy though. Had to make it because many trackers, including google fit, couldn't detect 200bpm. https://github.com/SMUsamaShah/heart-rate
https://github.com/Opfour/warfare - A modern HTML5 remake of Warfare 1.0 (1995) by Carric Moor Games. Turn-based hex strategy with city management, unit recruitment, tactical combat, and AI opponents — all running in the browser with zero dependencies. Playable but still building in additional features
https://github.com/Opfour/coeus-ci - Named after Coeus (SEE-us), the 100 eyed Greek Titan of the inquisitive mind — whose name literally means "querying." CI stands for Competitive Intelligence. A business intelligence OSINT tool that builds company profiles from free public data sources. Give it a domain — get back a scored report covering stability, growth, tech maturity, financial health, security posture, and transparency.
https://github.com/Opfour/op4 - Op4 is a terminal-based encrypted messaging application written in Rust. It provides end-to-end encrypted private messaging with post-quantum cryptography, routed entirely through the Tor network so that neither the content of your messages nor your IP address is exposed to anyone — not even the person you are talking to.
I built several prediction models for the market price of gold and cryptocurrencies. But this is always more a fun task than serious stuff - you can't trade with it. Yes, I think it would be possible, but not w/o having the hardware and data resources. I also built a music player with an LLM agent, helping finding the right tracks.
Built an Android app to streamline the messaging web-apps I use to stay in touch with some contacts.
I already refrained from installing those services' apps due to privacy concerns.
https://gitlab.com/not_john/palpipe
transitioned from my OmniOutliner 3(!) based system last year finally to modern macOS and obsidian.
1 year later, with no js/ts skills at all, i got 10 custom plugins, several forks where i fixed bugs and some custom adapations, dozens of scripts and snippets and what not
Now obsidian works for me like I want, and still every day I use it Im still in awe
It is a bit hard to describe until I make a video but basically I seem to have changed the fundamental unit from a document to a chunk and it is just awesome
I downloaded the database of traditional Irish tunes from thesession.org and now I have a tool that lets me search with fuzzy search and instantly pull up full screen renderings of every setting for each tune. I can also add tags and set my preferred setting.
It's really removed all the pain points from my practice hours, especially when internet access is shaky.
I made a timesheet entry, invoicing, and basic bookkeeping system for my freelance business. It works pretty well, I used "spec driven development" with Codex and it one-shot the entire application except for the PDF invoice layout which needed iteration.
The most comprehensive tool I've built so far was a JSONPath playground [0] when I was working on a game where I found myself writing a fair bit of JSONPath (well, JSONPath Plus specifically) expressions by hand and wanted to be able to test them out in the same way that I would on regex101 when writing regular expressions.
I realistically probably would have only saved myself less than an hour of crafting those expressions if the tool had already existed (with this level of detail, there are lots of many simpler ones already for it), but I would have spent a solid 40-50 hours of bouncing between manually crafting and writing detailed instructions to direct the agents to get this tool there.
Mostly I was not an "I have a script for that" guy before AI, except for the occasional VERY simple or VERY high ROI thing. Now I've got more oddball scripts that do things I couldn't be bothered to automate before, including: workarounds for buggy software, stitching tmux together with other tools, automating common operations so I don't have to use the god awful GitHub UI, etc.
I wrote my own Claude Code agent. It leverages some parts of multiple different agent stacks but using my architecture. It's layered, has long-term memory, agency to expand upon work, access to linear, access to all the models and endpoints you could possibly wish, as well as support for combining multiple providers and models into the layers so that delegation to the workers happen on the cheaper, more cost effective models. All of this runs at scale in an orchestration platform I wrote, using said agent, to create a cluster of docker containers in a JBYOS configuration, cross-cloud, k8s or swarm or whatever can run docker. It's pretty sweet and is basically an AI software development shop. I only have to give it ideas and goals.
As the 36 millionth person building something similar, I wish[0] there were better info out there on what works well. I can understand why, but it's still frustrating seeing on the one hand how deeply helpful the flywheel of this type of setup can be, and on the other hand how every blog post stops at some incredibly superficial setup to help them write more blog posts.
[0] Yes this is a plea, if anyone has the good stuff
Here's what I did, roughly, YMMV. I have a lot of experience in managing scale and web platforms as it's what I did for like a decade so, grain of salt and all. Let's assume you have access to N number of linux machines with GPU's, for the sake of argument.
I have a small RPI acting as my homelab pihole and dns so what better than to run the management UI on?! So I wrote a small bun management plane, nothing fancy, just a react app with user auth + openidconnect for those that like that stuff. From there, you have compute pool (empty at the moment because it requires a deployed agent). I added the ability to directly ssh into a machine, install the "agent" with privilege so it can manage docker, and the agent talks back to the management plane over websockets. A keep alive / health / status / resource packet every 15 seconds. Streams if you are looking at logs or accessing a container. I used Codex for most of this work but defined the protocol and everything upfront using protobuf (even though it's websockets). It helped with the "vision" and keeping the agent like Codex on the rails through completion.
Once you have a pool (agents installed on your N number of linux machines), you can deploy apps (which are my way of saying, a container with a namespace) or you can deploy agents (which is my agent, custom made for this) that are assigned to a project. I decided Org structures are a great way to delegate workloads so that's how they are modeled. Projects provide the git repo, the docker registry for images and storage of artifacts, as well as the history of all the prompts the agents have done in the project. Useful if you want to go back and search through |thinking| tags to figure out the reasoning behind a decision.
All of this was built in like maybe a month with Codex initially, until my agent was up to the task of coding w/ an endpoint configured (OpenAI API initially, now, NVidia DGX Sparks). What really works well is the delegation. The agent's have a webui that is exposed via the project urls so you can interact with the "scrum masters" of each project. They also share a stream if they are on the same project (but different subprojects).
I too wish there was more information on this but I didn't keep the lack of it from stopping me experimenting and finding what works. I came from the Mesos/DCOS era where you stop thinking about the metal and think in pools of resources. It's a distributed systems problem.
https://truepb.net/ - My unashamedly vibe coded strava analysis & athletics results site. I'm no dev but I've used it to up my knowledge in cicd, security, postgres and frontend/backend development.
1) a combination of Python scraper and Claude skill to help family members find job by matching jobs to resume, to rank best fits
2) similar to above, but for stock data and financial news to identify movers and why they moved and see if anything is interesting.
3) a couple of attempts to import EPIC medical data from hospital into an offline app. Needs more work, data export from EPIC is crappy and a mix of images, pdfs, text, HTML and .jsonb files. Not useful at the moment
4) an application that downloads stock market data to run 15-20 strategies and back testing to identify stocks that match multiple and then run sentiment analysis on news feeds. Interesting, but semi useful. Needs lot more work.
Blocks the computer for x minutes. Agents keep on working. AI doesn't need a break, but I do. And honestly, at least for me AI has made my desktop as addicting my social media feed.
This is awesome idea. Thanks for sharing. I extensively track my ScreenTime cross devices (phone, laptop) and try to reduce it under 6 hours a day. This would be helpful!
I made https://github.com/gagarwal304/meridian to analyze claude's open telemtry data and learn how to improve my claude.md for token efficiency and better output from claude code
A terminal-native agent multiplexer built on tmux. Similar idea as herdr but wraps tmux in an outer TUI instead of replacing it entirely: https://hmx.dev
I started a new job that requires a much more locked-down laptop than I'm used to. No macOS App Store, no random 3rd party apps. I miss a lot of the convince apps I used to use HOWEVER we have unlimited access to Claude CLI. I've been able to mostly one-shot working replacements for Alfred and a meeting reminder app called `In Your Face`.
My own Alfred replacement is actually better for me, it's tailored to exactly how I want to use an app launcher / shortcut tool.
A Linux dockbar with tons of applets and support for x11 and Wayland. Works on Gnome and KDE. I always wanted to write one as I have been involved with several open source ones, but it is a lot of time to go from scratch. I use it everyday, and I am enjoying it so much!
A git worktree shell utility to quickly switch/manage git worktrees, and a neovim telescope picker which switches all the loaded buffers into the worktree version: https://github.com/jasonwoodland/wt
A neovim session management shell utility (and telescope picker) to switch neovim sessions—no need for tmux or screen anymore: https://github.com/jasonwoodland/nvs
It uses an LLM to generate domain name alternatives that are relevant to your keywords, then checks whether any of those alternatives (in several TLDs) are available to be registered.
Warning: It's still a bit glitchy as I haven't fixed all the issues yet. It uses LLMs, but it's not a vibe-coded app itself. If it seems to be stuck while finding domain names, just refresh the page.
I've made 3 that use pi as the primary interaction, with custom tools and scripts.
1. family tree based on wikitree format. Transcribe records, verify/edit, then incorporate them into the tree with full citations and biographies. This one is the big one. It includes a tree browser and best genealogical practices.
2. Pool Math replacement. Log pool chemistry tests to markdown files and suggest the right amount of chemicals to balance.
3. Calorie counter. Log calories to a markdown file, look up foods and amounts in online databases, sync with garmin connect for exercise calories.
All of these are written with AI but also are interacting with pi and telegram, mostly using deepseek v4 flash.
I built a complete clone if fly.io infrastructure: VMs, networks, etc. so I have my Vercel on bare metal machines to maintain full-stack apps on https://playcode.io.
It took more than a year.
Why? Cloud infrastructure is too expensive.
TONS of tools. Most written in Go. Several have API servers in addition to cli or TUI or web interfaces. The API interface to my apps makes LLM-driven development much faster.
https://github.com/michaelteter/docgen : create a single text file of your entire project, with a tree and some other useful bits. This is good for dropping into an LLM or research notebook instead of giving an LLM access to your actual project folder. It also can be put in your pre-commit script so you always have one single doc you can diff from one commit to the next.
md2pdf: markdown to PDF, relying on defaults and optional config files or cli args for formatting choices (such as page margins)
md2gslides: markdown, converted into slides, and using Google Slides API to generate the doc in my Google Drive. This saves me so much effort (I teach, so I make lesson plans/presentations all the time).
get-music: TUI app that lets me search Youtube and easily queue up to download one or more of the search results. Then I take the downloaded content, split out the audio, LLM process the video title, add metadata for music, and then provide an easy command interface for local searches and playback of downloaded content.
bookmarks: TUI for slurping all the URLs from my browser, LLM-tagging each url based on the tag list I provide with the prompt and url, and lots of features for managing priority, show/hide tags, etc. This was to help me stop worrying about having a hundred tabs open. Now I can just sweep them up into my own private, encrypted (sqlite) db.
ESL-Planner: Complete web app for building class plans for teaching English (based on params, such as student age range, skill level, specific teaching language (what we want to teach), etc. It's close to being ready to productize and release as SaaS, but I built it for myself initially.
Numerous other tools plus a guide doc listing all the tools and what they do. These resources are then made available to LLMs when I'm developing, saving me (and the LLM) the time of hand-crafting the same tooling over and over.
Nothing that I couldn't have done myself, but in a fraction of a time.
- Custom off-brand version of Pangolin
- Dashboard with beautiful UI for parsing traefik logs with database, filters, map and various integrations and statistics
- Samsung SmartThings Volume Control for Soundbar in Windows 11 native style
- Android App with good UI which serves as remote for switching display output modes for PC for movie night / gaming night with various toggles and for remote game streaming
- Many little one-prompt apps which run in background for QoL
- Reverse Engineering with IDA became a walk in a park
A dashboard to see what my local commercial free radio station (89.3 The Current) in Minnesota is playing. It shows how often tracks are played, track and artist play history as well as some other fun lookups and visualizations.
This is cool! I was sad to see that kexplorer.org site no longer exists. :( I've been thinking about getting something like this up for a small Seattle station space101fm.org.
https://app.refutu.com - Tool for planning and executing together as a team, either at home or work, planning with or without ai and controlling the board with ai chat or dnd. Simple Goal -> Action -> Tasks. Paid for 3 or more users.
I made an android app for my badminton club which allows me to take registered players from a whatsapp group through to Square for easy payments and reconciliation.
* Spoof GPS - I live in a non-english speaking country. Sometimes Google sets my location to my gps, despite having an english vpn. This is an attempt to correct that.
* Local translate, rather than sending everything to Google.
* (non browser) An SSH selection screen, so I don't need to remember the IPs
I made a quick web app that lets me easily perform Bayesian evidence updates for a set of competing hypotheses. You drop your hypotheses into rows. Then for each piece of evidence, you add a column and fill its values with the odds of observing the evidence given that the corresponding hypotheses are true. The app then computes the posterior odds on the competing hypotheses, given the complete set of evidence. You can also import/export your results as CSV data.
A tmux config to handle my project based agent workflow
> agentmux
> Configurable tmux agent launcher. Define AI agents (or any CLI) in TOML; sessions auto-launch the correct agent, tabs are colour-coded per agent, and prefix m cycles through the list.
A catalogue for my e-books with integrated semantic search. I embedded the e-book records, saved the embeddings into a vector database and I can now search them with natural language: https://trnq.eu/en/projects/eblioteque/
I made a golang socks5 proxy that routes traffic to different VPSes (or the default gateway) based on hostname, over mutual-TLS tunnels, authenticated using ed25519 keypairs shared out-of-band. The "client" and the "server" are the same piece of software, and there's a web UI for configuring the routes.
I made it to deal with internet censorship in the UK, where different sites have different optimal exit jurisdictions, and most sites work fine so I can avoid the extra hops where possible.
It also works well for video streaming sites with geofencing, since the geofence itself is usually implemented in the frontend rather than the CDN. So only the frontend traffic needs to be proxied, while the bulk CDN traffic doesn't need proxying at all.
Socks5 is the ideal layer for hostname-based routing, since the proxy can see DNS names without needing to sniff TLS SNI (which would be incompatible with ESNI/ECH etc.)
iirc it was basically all done in a single prompt, and I've been using it ever since. The only issue I've encountered is with WebRTC not working properly with some services. (Presumably it breaks the NAT holepunching process or something, I never diagnosed it)
Another project that isn't quite finished is a "universal" web video downloader that works by shimming the MSE APIs and remuxing the streamed segments back into a regular video file. The idea is that if you can watch it, you can save it - including but not limited to youtube videos. I started this one pre-AI but AI was a huge help with the container format wrangling.
I wrote myself a little CLI app for generating 2FA codes because I got tired of the hassle of opening my phone and typing them in. So now I can just do 'toof nas' and get a code for my Synology account in my system clipboard. It supports nicknames for accounts, in case I'm thinking of "nas", "synology", or the hostname of my nas.
It still needs a bunch of polish, but I use it a few times a week.
* A 'docker compose' to provision email, chat and documents for human-AI hybrid teams where you can take over AI's agent's credentials temporarily: https://github.com/vezzadev/roster
I made a home assistant esp32 device to toggle the light switch for me. Codex done all the software stuffs, I just plug it into mac, let codex "this is a esp32, this is how i connect it with a motor, make it xxx". And it works exactly as I want after 15 minutes.
I’m building an agent-native deployment PaaS. I’ve done a lot of DevOps work and I keep starting side projects, and deploying with AI agents has been one of the most painful parts.
https://probplanner.com/ - I never had the time to dedicate to building a Monte Carlo simulator for project estimates. It was always something I just couldn't justify given my short commute. I used this project to teach myself how to use Claude Code and Codex over last summer.
I've done a lot of little things in Emacs since. Just minor things to improve my workflows or build up Chief of Staff type information flows.
- Day logger quarterly goal management and daily goal tracking system with multiple checkins, voice transcript task dump, jibberish to apple reminders, daily recommendations based on activity and goal tracking and always on dashboard.
- Snubnosed mandarin app. Vibecoded anki and tinder-like character game for mandarin which allows new vocab to be added on the fly. Also accurate text to speech for tones.
- What did I learn? Tweet summarize that takes all favored tweets and assembles into weekly categories and allows deep research on certain topics.
https://runnem.com - something I use to easily get projects running again when I get back to them. It also helps the AI get at the logs of the running processes.
Ive made my own agentic IDE centered around worktrees and containerization. it allows me to run multiple development environments on my machine with each development server running in parallel, allowing me to spin up feature branches and test them instantly.
1. Built an agent memory tool since all agents and clis are dumb and don't remember anything. Instead of prepping 300 project files and Md files I just say:
Check sugar memory for the latest thing we were working on.
2. The second thing is when making changes across a large codebase agents are also dumb at figuring this out and also grep 300 things, using tons of tokens. Instead I say
Check RemembrallMCP to analyze the impact of the change.
A little utility for Windows called TaskbarIconOverlay that puts a custom icon on top of a taskbar item. I have many VSCodes running at once and it's hard to tell which is which: https://github.com/jlahijani/TaskbarIconOverlay
* Auto-Birthday - if you have a contact in your android contacts that has both a mobile number and a birthday in their contact info, you can choose to send them an automatic "Happy Birthday" message on their birthday at a specific time. Can do it with hundreds of contacts. Doesn't use hardly any battery or resources.
* Wrecker - stupid simple "throw a ball at a tower of bricks and try to clear the board" game. High score tables. Made in unity. High battery when in use, No battery use when not playing. Will use internet for high score data.
* GeoNote - Create Geo-fences to generate a notification when you enter a location with your custom text in the notification. My wife is always telling me, "Next time we're here, remind me to only order one piece of toast" or something like that, so I make a note, it pops up the next time we're there and we're both happy. Notes are stored locally. No internet access required. Uses Geo-Fencing which is more battery-friendly than always-on GPS access.
All my apps are free, very privacy-focused and as battery-friendly as possible.
No information leaves your device (other than the high score data in Wrecker).
You have to side-load my apps though. I'm not putting them on the Google Play Store. They're so annoying to deal with! OMG
Nothing says "I don't care about neurodivergence" more than someone complaining when I build tools and processes to help me interact with the world better.
llm-consortium: prompts multiple models in parallel, loops until confidence_threshold, and iteratively refines a response.
This was inspired by a karpathy tweet [0] and the prototype created using another tool of mine: The LLM Plugin Generator plugin (essentially a curated collection of plugins for simonws llm cli as a few-shot prompt)
The llm-model-gateway companion plugin lets you serve models from the LLM cli as a an openai API. This allows you to use saved consortiums in your various clients as if they where a regular model. Bringing massive parallel reasoning to any workflow.
It occured to me at some time that an collection of parallel LLMs was not really a consortium. A consortium is a group of organizations. A group of groups. To rectify this I added for actual consortiums, where each member of an llm-consortium can itself be a consortium of models. e.g.
llm consortium save cns-glm-n3 -m glm-5.1 -n 3 --arbiter mercury-2
llm consortium save cns-k2-n3 -m kimi-k2.6:3 --arbiter mercury-2
llm consortium save cns-meta-glm-k2 -m cns-k2-n3 -m cns-glm-n3 --arbiter cns-k2-n3
Yes, even the arbiter/judge can be comprised of a consortium of models, bringing parallel reasoning to the task of judging parallel reasoning chains.
Consortiums can also now contain groups of specialists. These custom user-defined expert characters address the prompt from a different perspective. And a Westworld style Attribute matrix can be randomized to inject some more entropy into the process.
classifai
generates labels with approximate confidence derived from logprobs
llm-alias-options
saves inference parameters such as reasoning effort with a model alias. (good for setting the provider in openrouter or creating a consortium of high temperature models)
llm-prompt-json
adds a --json flag to return the llm logs object (good for getting conversion_id, or reasoning output in scripts)
llm-jina adds support for all jina AI specialised models and tools like web fetching, embedding and reranking.
- little visual web app cataloguing a small vineyard by vine vigour
- 5 min web app that helps show my daughter netball zones per position
- web app to track manual irrigation runs of 30ish taps across our property
- a calculator to cost-compare types of retaining walls by length
- a virtual card deck for testing a game idea
- scripts to help clean up an unruly Gmail inbox
All kinds of random stuff really, but to filter it down to only the noteworthy ones:
Tuber[0] - this is my favorite, use it multiple times a week. It's just a little CLI wrapper around yt-dlp for my most common use cases - downloading the video, or the audio, or the subs. And then, if you've got the Claude CLI installed, it can also shoot the subs through Claude for a summary. I use it all the time, it's a great little thing!
Scrapio[1] - this is really specific but I was so pleased with how it turned out. You give it a list of "hacks" ("mods" for Super Mario World) and it goes out to SMWCentral, grabs each of the patch files, and patches a clean ROM. I think I only used it twice but it was just a nice way to chew through a list of hacks and get a few ROMs ready, made quick work of something that would have otherwise taken a bunch of annoying schlep work.
Lotus Eater[2] - calling this a tool is a bit undersell, but I'm still really pleased with it. It's a fan site for jamtronica greats Lotus that scrapes Nugs.net for setlist data and lets you do some mildly interesting analysis on things like song frequency and co-occurrence. Also has a per-user "shows I've attended" thing, Setlist Bingo. It's been fun to hack on.
Lastly, less a tool, more just a toy: last week Google released their Magenta model for doing live music generation. I thought it was really neat, and it's open source, so I opened it up with Claude, and after a few passes and some extremely annoying toolchain issues, I was able to add a spectrograph which does key / chord analysis to the "Collider" app, so you get a live readout of "what the band is playing" and you can pull out your guitar or whatever and join the jam with some info at your disposal. It's the kind of thing that would have taken way too much effort to be worthwhile in the past, but with AI, it's a really neat result of a fun night of weekend hacking. See the README I added in my fork for a screenshot:[3]
I built an offline background remover website that now includes passport photo editor, object remover, image compressor. everything is free and offline (inferred from WebGPU) and I used to have to browse to different website to do all of these. now i can just do it offline on my own site.
- https://github.com/sethdeckard/atria : TUI for managing multiple AI coding agents that doesn't force a particular workflow on you. You can use tmux, the built-in PTY, or terminal integrations with iTerm2, kitty, or WezTerm.
- https://github.com/sethdeckard/atlas : TUI and CLI (also works as "cd launcher") that creates a smart, automatic map of every Git repository under your projects root.
This one converts a basic chunk of OpenStreetMap data to an SVG so I can mark it up (by hand) in Adobe Illustrator to make specifically-styled print/PDF maps, such as what get installed at trailheads: https://github.com/c0nsumer/osm_to_ai
And all of this has been put together to make the custom, local, specific-use-case maps that are at https://trailmaps.app (which, via local curation, are overall better mobile/online maps than many of the bigger auto-generated systems such as Trailforks, Gaia, RideWithGPS, etc, for visualizing local systems).
It's neat stuff where I understand all the inputs, outputs, and how most of it works, but AI tooling (Claude, mostly) has allowed me to bolt it together much faster than I would have writing it myself.
The idea came from using the Strava heatmap in JOSM to trace the proper location of mountain bike trails. I'm trying to use Strava less, and usually have ridden the trails enough myself before mapping them that I could use my own routes... So I figured why not have my own heatmap tile server?
It's also cool to just look at.
I could take it a lot further with time boxing what's displayed and whatnot, but generating the tiles is computationally expensive, so I just stuck with what I have for now. It meets the need.
I built a half-baked CRM that has a lot of custom fields and visuals for statistics that are relevant to my potential customers. I'm selling primarily to registered data brokers, so being able to pull up their self-published compliance stats (gleaned from their own privacy pages or public filings) and contextualize them in terms of the rest of the industry ("your deletion request volume has been in the 95th percentile year over year") has been extremely helpful when starting conversations. I also gamified it a bit by giving myself targets for cold outreach and gathering hard numbers on my cadence for outbound calls and emails per lead.
I also built this site for educating potential customers and other privacy professionals about the increasing tempo of CCPA enforcement actions driving compliance: https://ccpa.world/enforcement
I could have probably coded this from scratch quicker considering that it took me two weeks to remove all of the hallucinated imaginary enforcement actions against real companies and also the citations to non-existent California law that the models kept injecting into my enforcement summaries.
Tubenote, a free YouTube video summarization extension.
Mangata, a walking app that makes it easy to take notes and photos while walking.
NotebookLM Clipper, a browser extension for importing content into NotebookLM.
Knock, a notification tool that sends me a Telegram alert when Claude or Codex finishes a task.
I started something similar but for Sydney with the ultimate goal of being an agent simulator, simulating everyone going to work etc. Project got a bit unwieldy and is on hold while I mull it's future over.
A Few:
- Augsentric [https://www.augsentric.com] - probably my biggest time/AI sink - for evaluating websites
- FencePost - [https://github.com/seriocomic/FencePost] - a UI for multi-host Firewall rules (UFW)
- EventFeed (private repo) - a timeline of events on my network in a centralized UI
- Ledger (private repo) - personalized finance ledger using bank statements
The biggest unblock remains the tools/scripts/skills for documentation (started with Notion, network now sits on Obsidian for read/write).
I've been trying to make a comprehensive trading platform for crypto - with different verticals like DeFi and CEX. Why so? Because there are more libraries like ccxt to get data to analyse - rather than for the Forex and funds
1. A dashboard that tracks my personal metrics (github, strava, todo completion, flossing)
2. A eink display for that dashboard
3. A realtime node graph that shows a codebase (and/or its diffs) in a way that I can visualize what functions call which, and under what conditions
4. A agent that automatically fills out government forms and creates invoices for my friends brewery based on the delivery notes in their google calendar.
That sounds useful. I went a smaller/specific route with OpenHop: not a full graph of everything, just agent-authored flow walkthroughs you can step through locally.
I host a couple of services on a box in my network. I built a tunnel that runs on the box and on my VPS to allow me to quickly access those services.
A tool that checks for new movie and tv releases, looks up ratings to see if they are worth adding to my plex server (see above about services I'm running), and then finds the magnet link and downloads them. But will only do so if my VPN is connected.
A tool that allows me to quickly build out paintball fields using my STLs of bunkers that I made, and export the full field layout as a single STL for quick painting, slicing, and printing.
replaced some paid apps with local - google reader rss replacement and send web to kindle.
most ambitious was browser extension to automate booking reservations since captcha and timing were critical.
swamp/evaporative cooler controller that monitors rate of change of relative humidity and sends RF signal to cooler to turn water on for x seconds to keep humidity lower than just leaving set to cool all the time. Does a great job but need to work on edge cases and ultimately replacing remote with my controller.
financial models for retirement planning
pen plotter gcode creator for old 3d printer to make labels.
food monitor to track what i ate with AI analysis based on symptoms recorded as well as interactive follow up questions
inventory app for electronic parts that now is inventory for all kinds of stuff like tools/plumbing etc that uses photo/ai to fill in data.
I like the capabilities of C++ and imgui but didn't want to deal with C++ anymore so I had AI do it.
imping - PingPlotter-like app. They didn't have a Linux version and I'm a paying customer, so I vibe coded this one:
https://github.com/zenakuten/ImPing
There are plenty of other tools already that do this but hopefully this one adds some quality of life niceties such as dark mode and a mobile responsive design that seems nicer to use than the others out there.
Hope the colours look ok. I'm a colour blind person (deuteranomaly) and I've optimised it for what looks good to me. Am open to adding a mode in line with what regular vision people might prefer.
No frameworks used, vanilla JS. No sign ups. Data/state persistence is purely local storage.
Absolutely no commercial interest here. Built this purely for the love of it. Will forever be free.
in case you forgot instruct your ai text generating tool to do so: replaces — with -, removes emojis and changes quotes to look like human-typed (even though they are not grammatically correct)
Been working on this spaced repetition app / incremental reading app. I used SuperMemo for about a year straight, found the UI not that great, so I decided to make my own incremental reading app. Think: incremental everything. Supports multiple spaced repetition algorithms FSRS-6, SuperMemo 18, SuperMemo 20. https://github.com/melpomenex/incrementum-taurihttps://readsync.org
I don't do a ton of programming on my own time to be honest. But I did make some recipes with Calibre. Seeing Gemini basically jsut take the HTML from some news Web sites and come up with recipes really opened my eyes to how good that stuff had gotten.
https://termonmac.com/
A relay terminal that lets you connect back to your Mac from your iOS device.
I spent about 2 months building TermOnMac.
I am going to develop the next version TermOnHost, which will let you connect to all kinds of hosts (Mac, Windows, Linux, or any Linux embedded system)
And your hosts can connect to each other.
I've been working on a tool to design laser cut jigsaw puzzles https://jiglu.dev/
Since I live far away from family I also added an online game so I could play with them or show them what I was making more concretely.
I've cut some jigsaw puzzles that it made, but without access to an uv printer or a laser cutter that works reliably it's been challenging to actually make them
HN Chrome Extension for dark mode and a few other styling tweaks I wanted.
Jira Chrome Extension to add some notes and links to dashboard pages that I wanted for ease of use.
Small application which takes a CSV and turns it into a Registered Server List for SSMS in order to keep my list of servers updated for queries across all our databases.
Honestly one of the good use cases for AI. Small low complexity scripts and tools for assistance is a great use case. I'm amazed at the folks that are doing huge monolith rewrites with Agents and such, but I've never had good results for that. Small time saving scripts gets me a much more direct return on investment.
I built an RSS feed reader for Claude, https://clawfeeds.com/ it does very be little: 1. check the feeds, and 2. Turn the output into easy to parse markdown
I've made text editor engine for .NET 10 built in C# with no UI dependencies. Includes a piece-table buffer, multi cursor editing, syntax highlighting, code folding, snippets, TextMate grammars, diff engine, undo/redo, C# REPL, plugin system. Its essentially a deterministic execution runtime for structured, auditable text mutations thats built for AI systems, automation pipelines, and tooling that needs to manipulate text documents.
https://github.com/marinusmaurice/TextApi
So many! I manage a fund that buys small e-comm brands, and at this point the whole thing runs on a combination of AI and tools created with AI. My favorite is one that scrapes my Alibaba/WeChat/WhatsApp/email supplier convos daily and uses that to build a dashboard tracking the status of my orders.
Most of it has been to maximize productivity with AI
1) Use chatgpt pro from codex cli, opencode, claude etc as you can't get it via API. This has been the biggest boost in productivity for me as I don't have to copy and paste.
I've been building a tool aimed at better web annotations for teams and AI collaboration at https://viewall.io/
Having worked with web facing teams there are always vastly different methods any individual uses to capture their feedback. If you or anyone you know on Mac that has 100s of screenshots on their desktop, this is aimed at bridging the gap.
Clipboards are optimized with context for LLM markdown ingestion and for use in work suites like Jira/Confluence.
Still fairly early, but I've been using the tool to help build the app itself which has been an enlightening experience.
Hey brad. Love your blog's layout. The neon sign headings are so catchy! And I'm going through some tough times in my life, and I can totally relate how cathartic the process of 'creation' is. Thanks for sharing!
I finally shipped a Chinese learning app I wanted for myself for ages at https://wenmoji.com/. Just never had enough time to sit down and code it end to end. Still need some improvements of course, but will slowly chip away on it. I use it daily/weekly myself now.
Highlights are that it completely free, no login required and works offline (once you "downloaded"/cached some files the first time around).
I also have a local zsh autocomplete macro that let's me type things like "git rename annotated branch" and ctrl-g and it will get me the actual command. There's also a ctrl-r mode that searches my history using natural language. This is connected to a locally run ollama so my keys don't leak.
- ticketing system with Stripe payments and QR scanning at the door
- Instagram/media ingestion for the club site
- genealogy tool with GEDCOM import
- scripts for downloading/archiving public-domain film material
- playlist/library tooling for DJ use
- music collaboration/sync tool for Ableton projects
- normal work stuff in a much larger existing codebase
I have become a lot more strict about process after being burned a few times. Mostly: make the change small, be clear about what it is supposed to do, check the assumptions before coding, use tests/logging/manual checks as evidence, and don’t merge anything I can’t review and explain myself.
Wispr flow released android version few months back but wasn't supported in my 6 year old phone so made a similar app named Floatspeak. Which motivated me to made a windows version of it and now stopped paying Wispr flow altogether.
I made a Safari extension with Swift that automatically suggests using Fastmail masked email addresses on login forms. Never published it, instead just using an xcode dev build on my phone. Works flawlessly.
I am building my self-hosting llm-wiki system (https://gist.github.com/karpathy/442a6bf555914893e9891c11519...). My approach is to start with a theory of how such systems could work. Then since llms can interpret theory - this theory becomes an executable llm-wiki system itself.
here are a few i have put time and effort into. these are not “vibe coded”, but an agent was utilized at points to save copious amounts of time implementing my architectural decisions; my schedule is pretty slammed as is.
https://mithraeum.studio - local first agent and editor in C, also a few models on HF (mainly jsut qwen wrapped atm but working on from scratch)
https://fieldopt.dev - SaaS for dispatching jobs to the field (technicians, trades, delivery, etc.)
https://github.com/zblauser/ytcli - youtube music from the terminal in zig (ps it’s free, no sub needed)
A github client / dashboard that can pull 20 of so repos for all internal and client projects in one UI so I can stay on top of project delivery and long standing bugs. It has global search, bookmarking and text based / minimalistic ui for maxium space utilisation and information density. It's read only so to comment on issue i click a link to open GH in new tab but helped me a lot to have this birds eye view on my company. Don't get me started on GH Project. I tried Linear many times but multi project / multi repo is just not their core focus and it shows.
A Linux kernel module and userspace app to read the performance level of my Corsair AI Workstation PC (Strix Halo). Corsair only ships a Windows on screen display app and I was flying blind. AI helped me look over the Windows installer, decompile ACPI tables, and identify the WMI events. Built it all in Rust. When done I asked a different AI model to perform a security assessment and help me harden it. Way out of my wheelhouse, yet an afternoon project with AI.
I made a media center replacement for something like plex or jellyfin, streaming video or audio whenever I am; transcoding, subtitles, specialized dupe and renaming metadata. A little automated datawarehouse that manages all my output in an object store. My own tag system of course. A personalized eval system for llm tools.
I've made a harness to discipline it and get consistent output regardless of model. Using it daily. Is the opposite of vibe coding, it delivers great planed code with my engineering taste. I had it open sourced for a while then I've closed it. Just a month or two after closing it, I read an article about this "clean room" thing https://en.wikipedia.org/wiki/Clean-room_design
Virtdev, my own rootless development virtual machine system. Use it every day. Even integrated it with tmux.
Pugneum, my static site generator based on pug/jade. Technically made many years before LLMs, but AI is fully maintaining it now so I think it counts. It's gotten to the point I believe it's superior to markdown.
The team behind Gel got acquired by Vercel and I already experienced falling in love with a dead database (RIP RethinkDB) so I decided to fork the concept to a TypeScript port with Svelte as the UI instead of React.
It'll live at disc.sh in a few months. Early dogfooding is promising.
https://github.com/haydenk/homestead - another Go project, working on a better uptime dashboard that can also be the main homepage for accessing homelab resources.
I also used AI to find and create issues and milestones to for me to get the project to 1.0.0
It evolved out of some weird interaction someone was smartassing me that the moon wasn't full when I was pointing to how pretty the full moon was. After that, between a friend and myself, it became a bit of a running gag how full (or not full) the moon actually was. This was my first real project I kind of "vibe coded": https://moon.masca.teide.cloud/ - showing you how full the moon is to the 10th decimal
Scrobble tui to track vinyl record listens on last.fm, sourced from collection on Discogs
Calibre web UI 2.0 to replace Calibre's mediocre web interface. Used for browsing ebook catalog, searching and cataloging with a simple list feature.
Project Gutenberg local mirror UI to browse my local copy of Project Gutenberg books
A couple of MCP servers for self-hosted services to give access to OpenClaw. Currently working on a daily digest that Claw will generate that includes feeds from these: what news stories were popular in my feed reader, did my baseball team win, etc.
I built a terminal app for myself that conflates worktrees & tabs, runs every pane through a terminal multiplexer, lets me join in from my phone and generally makes me happy.
Could we please stop putting price tags on 15-commit repos?
It's just crazy that every idea, created with ai, now costs 10$ or more per month, despite it costs 5$ to create.
html snippet playground - for testing html/react snippets
token speed calculator - for estimating tg/s of ai based on ram speed and model size/params this helps in comparing different hw, estimating likely speeds i will get on hardware
prompt assembler - to create prompt and context once and reuse it in different ai's, picking and choosing context in a prompt, creating agent.md etc.
dashboard builder - for viewing gsc, ga, stripe data in one place
- app to help buy/find books for my wife
- app to help manage my climbing wall
- app to help finding good films/series
- app to track weight
- app to manage my board games and find the right ones to play
- app for planning wood builds (e.g. climbing volumes)
- telegram bots for:
- picking restaurants for weekly lunch with friends
- managing our 5-a-side football games, make teams, elo ladder
- fantasy football leagues
Tools? Mmmm i tried making a openCV based tool to recognise circular objects on sat. imagery, it didnt work at all (tried for two weeks), changed to a LLM, it seemed to work, or it was just a little bit off, O gave up. If anyone wants to collab in this space, shout out! Ah also a translation tool for pdfs… its also stuck in limbo, forgot why again…
I built a browser extension to create podcasts from HN stories in French (and English), I created it for myself in first, then I released it with a shared quota for the community but no one else uses it as it was forbidden for me to post show hn.
Made a local, in-memory OTLP traces viewer because I don't like running heavy alternatives like Jaeger, Tempo for simpler use cases https://github.com/pawanjay176/trace-top
I vibe coded two apps, I program but not in the languages used in the apps. I'm utterly clueless about Lua and Python. Both apps work fine and I use them all the time myself.
1. A Mecrisp-Stellaris Forth LSP for the Helix editor (Python)
2. A CMSIS-SVD Sqlite3 search and paste pop-up window for Neovim when editing STM32xx embedded Forth programs. (Lua)
In 2023 I wrote from scratch a iOS native app using SwiftUI. This year I used AI extensively to improve and add many features to it in a span of couple months. The app is free and there are at least 2 users of the app - myself and one complete stranger (not a family or a friend) that is using this app.
Also I had a redis clone from before, but with AI i separated the epoll layer from the actual database engine which made it sort of embeddable.
And hooked that embeddable database into a JNI interface, and now it can run inside Android applications, sort of like a concurrent hash map, but one that lives off heap and has support for TTLs and mget, msel, mdel, hashes and all that.
I use it for some silly android games.
A weight lifting app. I’ve paid for, and used, others over the years, but I always wanted to customize them in some way just for me. So, I just decided to create one the other day (used antigravity CLI) and I’m hosting it on Vercel as a PWA. I’m enjoying it so far and see a lot of potential with making hyper-personal software moving forward.
Dictation tool which works better than the built-in Apple functionality, for my use cases.
Bc my version uses simple copy/paste rather than deeper OS integration, it works more reliably in the Claude Code terminal (has to do with active windows or cursors or something). And bc it uses local Whisper, I also find it transcribes technical words such as “git” more accurately.
Nothing technically challenging but practical for me.
Working on a web client and bouncer for Hotline, the old chat software from ~1997. Just want to chat with my peeps who still use Hotline, from the comfort of the browser I have on whatever machine, while some server maintains the persistent connection to the Hotline server for me. Like an IRC bouncer, but for Hotline.
Built a quant system that reads earnings transcripts for what management is trying not to say. The model is surprisingly bad at this. Turns out management is too.
Starting making hyprwhspr because no other stt library was quite there for performance and model availability.
After that I started writing opub.dev because even minimal success in recent oss showed me just how much has changed, and I’m worried about how expensive everything will get for maintainers.
So, now I’m trying to GIVE people compute so they can start building a helpful filter layer above their projects.
After coming back from paternity leave, I found that my team had really leaned in to AI driven development. This project was half catching up and half attempting to solve the burn out from the repeated books my wife and I were experiencing.
I'm building an app that uses cosign similarity across a bunch of vectors to derive team productivity metrics. To be honest the maths is trivial; the hardest part is gathering data and normalizing it in a vaguely sensible way.
I've also built a release notes app for my QA teams, a DORA metrics app, a thing to map UX journeys with Playwright, and a ton of games and stuff. AI got me back into enjoying building things again.
I wrote a note taking app that synchronizes across iOS/iPad/MacOS and stores my notes in markdown files so that my agents can summarize them each morning, delivering me to-dos, etc.
LockIn - Beautiful scriptable terminal countdown timer that can block time waster apps. Enjoy fun visualizations and improved productivity that your agent can trigger to start a focus session. Install today with brew.
Built an Apple Watch app that streams music from Plex. It’s more stable than Spotify and Apple Music and it’s been a blast running to my own music collection!
I recently posted Show HN for https://www.useorganizer.com/ which helps organize stuff using timelines and stores data in a local folder not the cloud. Open source.
We used AI to build our AI platform and now we are using the AI platform to build the tools that we need for AI. :)
But no honestly, unfortunately most tools I did for myself are not for hobbies but something that I needed for work... like this one (https://github.com/crmkit/crmkit) most recently.
Oh man - I've created more than I can count at this point, but here's some of them:
- A chat-based web app for ad-hoc telemetry data visualization.
- A firefox-extension meeting transcriber
- A personal chief of staff
- A web-based personal finance budgeting tool (no AI at runtime, obviously)
- An iOS and Android app for solo work with a dead man switch if timely check-ins don't happen
- A custom dotfile/machine config manager that works the way *I* want
- A bookmarking tool/web clipper that puts together daily content-only collections from what I've been clipping and send them to my Kindle. This one I actually intended to make a SaaS, but meh.
Being able to put together quite big projects by myself for myself in a reasonable amount of time is such a joy.
Somewhat related - I wish there was some local thing I could give my 100 holiday videos and it made something fun with the highlights to a specified duration.
I'm close to releasing a memory safe programming language, with a declarative concurrency model, that runs on a Go-like runtime.
It has "levels" of compilation, with EASY mode being about as easy as Ruby, and the compiler can present you with options to get that as strict & performant as Rust/Tokio.
I'm going to need at least a month to finish all the documentation, though.
I'm building an app that generates lifting mesocycles and tracks every set and rep. Each week, it uses feedback from the previous workouts to adjust training volume and intensity. It's replacing an app I currently pay $25/month for.
A desktop markdown editor for design docs in git repos with markdown diff highlighting. Has been a time consuming but super fun experience https://github.com/emrul/md
I'm building a replacement for TablePlus: a TableAI database client, because the latest releases of TablePlus have gone down in terms of user interface quality. You can find it on the Mac App Store(TableAI - AI database client)
A few, but the one I use regularly and am quite proud of is
https://mediaden.ca - iOS app for storing encrypted photos/videos on storage I (the user) exclusively owns, with zero servers, zero telemetry, and a host of other privacy related features.
Recently I made Vocast (https://github.com/cnrmurphy/vocast) - a cli driven tool that uses local TTS models to convert articles to "podcasts" and expose them via RSS feed. I wanted a way to listen to articles without having to pay for an app. I convert the article on my PC (which gets added to a managed library), run a web server, run Tailscale on both PC and phone, then I can use a podcast app to access my library. Nice way to consume some articles while out for a walk or anything else and has worked reasonably well for me so far.
- a personal and private webpage for:
health: garmin metrics, apple health metrics, blood tests, rx..
- a kind of readitlater and bookmark index
- personal finance: wip
- in my homelab only available within tailscale.
The final idea is to own all my data, but I’m still on it.
I use agents to do most of the tedious admin for my hire business, and I built www.vessels.app to run them on the go because there was no native solution to talk to my agents. I’ve started working towards releasing this to the public because it’s so much better than setting up agents via telegram or slack.
I’m a UK teacher. I have built a custom GPT that marks essays for the subject I teach in a repeatable and reliable way. It gives actionable feedback to students.
I use it, and have given my students access to it too - they use it to help their revision.
Kids making mud cakes, climbing trees, and taking shots at each other with shanghais. Teachers lounging on a beach somewhere. Meanwhile in classrooms, bots.
How about critiquing? Like, the moral equivalent of an editor pointing out issues and/or suggesting alterations? I think it would still be the student doing all the things you pointed out, but I suspect there's a fair amount of leeway in the interpretation of each of those.
It essentially uses youtube as the music source, I think I heard somewhere that playing through embedded videos skips ads but I'm not really sure, in all my time testing it I never noticed ads, but I'm also on premium so that may have been why.
by all means critique, I don't know that I have a ton of time left for it and I'm sure there's bugs here and there. I was having issues getting it to autoplay on desktop when the window itself wasn't the active tab. I never really tried it on mobile.
I was trying to get some DB of artist/song info but doing that was proving to be complicated.
I'm working on a recommendation service (which, to me, it's the piece I'm missing when I play my local mp3 collection)
I collect song metadata from various places (genre, instruments, track credits, rating). I also scrape charts by year, genre etc.
Then I run an ETL job on the json data I have downloaded, pre-building queries for extremely fast lookup tables. This gets saved to Duckdb, which is used by my go web ui/api.
It's very early days, and I only spend one or two hours a week on it, but right now it's amazingly useful. It had roughly 80k song metadata. To preview the suggested songs I ended up building a very cut-down YouTube music player, except that the playing song has all the metadata right there, and everything is a link that can take you to the artist, composer, instrument, genre, album etc. It's a great way to "wander through your collection".
Unfortunately this is only useful to me, because I targeted the music I listen to.
Next step is to download lyrics and extract song meaning, keywords etc. Then use MusiCNN,
(or CLAP,OpenL3, HTSAT) to extract embeddings. Finally train my own model for
nearest-neighbor retrieval based on a mix of metadata, giving the user the ability to tune it on the fly.
Did you ever have to pass Appstore review process? How do they look at copyright and stuff when you are publishing an app that plays your local mp3 collection (how does your mp3 collections ends up on your phone?)
A ninteneon3ds game explorer where i can look at games and bookmark them with comfortable screenshot preview so i can check what i would like (i never knew what game i should play on it and there are hundreds)
i made a tool thats a combination of 2.5d cad and smart stylus for making things i can print for leather making, with embroidery patterns on top.
ive made some wallets, a incredible pair of ear muffs, and a bunch of key tags.
i keep being asked if im selling anything, and when i get the next piece together of building/buying an embroidery machine that can work on leather, i might
i still dont want to buy a proper fabric cad system, so im trying to figure out a minimal version for making glasses cases for everyone's christmas gifts. its handy being able to draw inputs for claude, but its also nails-on-a-chalkboard
Code review tool that breaks up diffs and regroups fragments based on runtime execution paths and/or architectural boundaries. I find it useful sometimes to see changes organised that way.
I was able to create a CLI (https://github.com/gitsense/gsc-cli) without knowing Go. Like 0% Go knowledge. It is currently over 300 files (266 Go files).
Sure! I wanted to mimic the process of daydreaming on a car ride, where external input and inner monologue can mix.
I wasn’t planning on posting it yet, so I’ll have to get you a thorough follow up, but in a nutshell it’s: a continuously running 448-concept space (philosophy, cognition, art, nature, math) that occasionally “crystallizes” a group of 4 concepts via Hebbian learning and stochastic noise. Those concepts get sent to the LLM with minimal guidance beyond some safety guardrails and encouragement to be creative. It has access to a sandbox to produce essays, stories, music, “art”, and small python scripts.
Self-updating memory system. Notes, essays, and artifacts can also be discussed with me through some outbox channels.
On top of that it’s got a separate academic philosopher + psychiatrist llm that critiques its work and has a regular cadence of “sessions” with it, as well as a research assistant bot who I talk to (but doesn’t interact with Sisuon) who has full project context and memory access. The sisuonspeaks site is a VERY abridged collection of Sisuon’s essays, along with analysis, commentary, forum posts, and a podcast…all created by, you guessed it, more LLMs.
oh boy, lots. i made a trainer that coaches you in nondual philosophy by quoting from the Upaniṣad; a Vedic Aspectarian that calculates your chart and analyses your transits; a better I Ching program that utilizes a time variable to throw the hexagrams; and then there’s our research software. none of this would have been possible without AI.
I wired up a stream deck to perform long-running tasks. Very much tailored to the kind of HCI that I prefer, so I can be interrupt driven versus checking on status all the time.
Eg: push a button, it shows that it's working for a while, then strongly flashes when it's done (success/failure). When you have it right under the monitor, it's like a macro pad for long-running things.
This reminds me of some of the very early peripherals you'd see on the Alto and other computers. I was surprised something like this didn't seem to exist, but maybe I'm just terrible at searching.
https://charleswiltgen.github.io/Axiom/ – Suite of skills, agents, and tools that make general SOTA models actually good at building and/or auditing iOS/macOS apps. Built for myself initially, I FOSS'd it once I determined how generally helpful it was. It's helped me learn a lot about doing sophisicated things with LLMs in a token-efficient way.
https://charleswiltgen.github.io/TagLib-Wasm/ – Also built for myself initially, I FOSS'd it because there was nothing like Mutagen for TypeScript/JavaScript runtimes. (I don't dislike Python, but think it's a bit of a mess.) This was my first serious project to leverage LLMs for coding.
https://pwascore.com/ – Built because I wanted to quantify how bad Safari was at PWAs. Learned that, objectively, Safari is as bad as PWAs as Firefox (which is to say, not terrible, and not to blame for why PWAs continue to be mostly-irrelevant).
basically trying to see what a vertically integrated agent looks like, where the agent has deep access inside a framework and it operates from within a framework, so like, instead of reading files, opening processes etc - it gets a bunch of framework specific runtime tools(logs are the easiest example)
I stopped paying for Wyze subscription after replacing the camera backend service. Saving me about $30/m and a much finer tuned OpenCV to Claude API vision model.
What were you getting for $30/month from Wyze? We have a package that is a flat $99 per year for a dozen cameras (plus however many more we want to add for no additional cost), unlimited recording, etc.
I wish I had time, but I would definitely make some Android apps to sideload onto my phone. They would be very bespoke and probably only relevant to me, but they would be streamlined to my life.
I built https://sdocs.dev and use it daily. It’s a CLI-driven markdown reader which (privately) renders Markdown in the browser.
When you install the CLI, it (with your permission) asks to update your base agent prompt files (e.g. `~/.codex/AGENTS.md`, or `~/.Claude/CLAUDE.md`) with info about how to use the tool.
This means all your agent chats know about SDocs, and it’s nearly always your agent which invokes the tool: “Hey Claude, sdoc me a list of all my open MRs”, etc.
www.propelcode.app - cursor on my phone.
www.propelagent.app - voice agent for my home health care agency, but it also tells bed time stories to me and my daughter a few times a month.
I also built a new web framework we use internally which is amazing. We might open source it soon. It has a postman clone that has a bunch of features I wanted. It really is the case that we can just build tools any time we want.
Looking into twitter right now, it seems like the way to do it is with headless browsers, but they usually cost money.
For reddit there used to be the json endpoints that you could just fetch, and you can batch your subreddits, so its nice end easy. They have just killed those...old.reddit still works, but i fear like the days are numbered there as well.
Reminder.dev - Quran app and API that includes RAG search to provide a more authenticated source of summarisation. The first thing I dabbled in with AI.
Micro.mu - Rebuilt my entire product idea from 10 years ago as a super app for daily digital habits. Something I use everyday for myself.
Aslam.org - An islamic knowledge base that uses sources of data as a way to ground AI chat and make notes. Very useful tool that I'm using on a daily basis.
Go-micro.dev - Totally revamped the open source project, docs, etc with Claude.
Micro.mu looks interesting, your description of daily digital habits made me think it was a productivity app, when it really is more of a dashboard for different channels of content on the internet (social, video, chat, etc.) what was your original vision? I am curios. Also I really like the design, its so clean, Ive never seen anything quite like it and yet it seems like such a good aesthetic that should have been obvious. kuddos.
Originally I felt there should be an alternative to Google and Facebook. It felt like our lives were being ruled by tech giants even ten years ago..and then all the advertising and algorithms really messed with us. I personally wanted an alternative way to use the web. So it's based around what I'd use on a daily basis. And then the idea of being able to extend it with a native app builder. The big stuff I continue to build with Go but ad-hoc things I can knock out with some JS in the app itself.
Mods! I've written mods for a couple of games that I've always wanted to but never had the time to learn the SDK for. Most recently for Project Zomboid.
A website that tracks when we last went on a weekend trip and other kinds of things, and reminds us with a cute friendly (not AI written) email when it's time to plan something again!
It really helps us to not forget to spend significant time with each other when life is busy.
A rant follows.
I've generated probably as many lines of code by this point as I've written myself over the past 5 years or so.
I found AI generated code mostly very frustrating, kind of low quality in its own way, and too complex. I have pages and pages of instructions to guide the agent(s) to do a better job at this, and it has gotten better, but the fundamental limit of this technology is tangible.
Like, okay, CPUs still get faster every year, and every now and then someone makes a breakthrough and we get a bump in speed from something. But when you write high performance code, you very quickly run into hardware constraints, like how fast information can move and how far away components are from the CPU cores themselves. People keep saying performance isn't THAT important, and that modern hardware is so fast and amazing, and that they struggle to even find a way to use all of their CPU cores and RAM with their little app or program or game. Yet here I am, writing code that will noticably speed up if I run it on a CPU with a little bit more L1 cache.
This is similar to how it feels to program with AI when you're reasonably competent (to put it mildly; I avoid the 10x developer label because comparison to others is very silly). Everyone keeps saying it's getting so much better, and it's so good, and worrying about code quality and architecture is dumb because we can move so fast it doesn't matter. Yet here I am, writing code by hand because I tried doing it with AI a couple times and it just doesn't hit the mark.
I'm not doing anything special, I just have high standards and a good amount of experience when it comes to software quality, performance, and maintainability, which is why I keep getting hired. I'm convinced that people who think their AI generated software is good are the same people who write short variable names and think it makes their software faster (hyperbole, but you get what I mean).
I can feel when I hit the limits of the hardware, and I can feel when I hit the limits of LLMs, and I know for both of them that a 2x increase in performance will not change what is and isn't physically possible.
I made a modal editor, like Vim and Helix, in Rust that has some prose features I wanted, and with a keybinding system I find more logical and consistent.
Started making Agentikus as a way to manage my -back in the day- multiple OpenClaws. Soon enough realized that many will have the same problem soon. And started adding features that I was missing on Codex and C. Code. It’s a fun ride.
Side note: before coding agents I would not get passed the branding and login page.
Well, I've been pretty active in our rec baseball team for the past few seasons, so: 1) App to help my son and other kids learn baseball IQ, and 2) Streaming app to compete against GameChanger.
It's been refreshing to say the least. :)
Forgot to mention: I built my personal finance app. I've been using a spreadsheet for almost 12 years now. Different versions over the years, but the same concept. It took me 3 days to get a web app that works better than the spreadsheet, and it's been life-changing!
i built a program that watches wifi traffic and if it sees my phone connected to the office wifi; it marks me as in the office on our internal chat tool (Zulip).
And the inverse as well, of course.
Runs on a raspberry pi that I was otherwise using to take backups periodically… has been working pretty good honestly.
I also built a program that fills/submits my time reports, and does the same for all of my subordinates - then signs them off… Saves everyone like 5 minutes if we remember to do it, or 12 minutes and frustration if we’re reminded by HR about it (which happens because who the hell cares about time reports?)
I've been reading Finnegans Wake and not making much sense of it so I wrote a data pipeline to scour the web for interpretation books / guides (which I also can't be bothered actually reading) and coupled it with an prompt to image pipeline. It's now a readable albeit silly picture book to go along with an audio book... I call it finnegans slop.
Thanks for the reminder, I made a very simple and small mac os guitar tuner and intonation app (all others are either paid, have in app purchases or are outdated) that I still need to publish!
CodeMerger: https://codemerger.nl
I've never liked the lack of control I feel using agents or tools like Cursor or Antigravity. I found myself having much better results simply pasting full source code in free chat, so I built a tool around that philosophy and I've been exclusively developing with it for a year now. It includes a Project Starter and code architecture analysis tool.
Presentable: a photo library sorter with Ai powered organization assistance, a compare canvas with various viewing modes and customizable folder sorting shortcut templates (wip).
AlwaysWhisper: a tool that let's me attach STT to my entire OS, with custom wrappers for different programs, adding theoretical voice control to any software (wip).
ScreenLoader: an Electron based tool that can load any web source as a kiosk app, full of useful features like keep-alive, covering multiple screens and tracking output logs.
Inputboard: a unified all-inputs hardware board, that transfers input data to any prototype I want to work on using an optocoupler, so I won't have to fiddle around with setting up clean and reliable inputs from cheap Chinese components every time I just want to test something.
Squire: an agentic board game helper, that can ingest a manual and will hopefully help decrease the time spent on endless discussions about seemingly conflicting rules. It should also be able to help me play a game when I don't fully understand the rules myself yet (wip).
NodeRunner: an agent that plays the WikiGame, focusing on speed, efficiency and token usage (the result of a fun competition with a colleague).
Sonic Bloom: more of an experiment than a tool. It's a wireless piece of custom hardware, that listens to conversations, sends data to an LLM through fast STT and returns a color choice that matches the topic being discussed to the hardware, which then controls an LED ring. It also has a small display that explains the logic behind the color choice.
Image-to-story: a VLM tool that kickstarts a written story using an image and has some rudimentary tools to expand on it based on user instructions (wip).
At-work-or-not: and Android app/website where colleagues can check if I'm working from home, if I'll be at the office or if I'm not working at all. Also doubles as a private record for tracking transport expenses.
SharedMaps: a Maps based website where groups of people can share custom categories of geo locations and drop comments on them.
VMG: an image format that includes audio with images and offers TTS input to easily add narration.
Who wants Coffee: a small Android app to help me remember who wants to drink what when I go for a round at the office.
And a pile of Python scripts for smaller useful tasks.
I bashed out a dashboard for myself the other month, monitors firewall alerts/warnings/shows connected devices, process monitors on a few pc's that I keep an eye on, a to-do list/calendar combination that let's me track some internal tasks I need to do weekly, bi-weekly, monthly, etc, oh and it pings cloudflare to keep me up to date on the website stats. And an Amidakuji game to help me and another staff member pick who's doing X task that day :P
Nothing major, and only works with my infra, but it saves me a few minutes a few times a day to just be able to check the tab, and if there's an alert load up the full stats page.
I get annoyed that existing tools have limitations so I fix them or build my own:
- I didn't like that I can't use my newsreader on my laptop and my phone as easily so I built https://github.com/mjc/nntp-proxy. that turned out to be really hard to benchmark once it got fast enough so I am working on an nntp benchmark tool https://github.com/mjc/nntpbench. both can do request queuing because the nntp RFC says servers have to accept as many requests as they can, and then process them in order. so if your client doesn't do that, you can use more connections to the proxy and it will queue for you. it also routes stateless commands to whatever server is least-loaded, and will switch to stateful mode if your client needs it.
- I didn't like how expensive AWS Transfer Family is, so I built this https://github.com/elixir-ssh/sftpd and then rewrote it in rust (alpha) https://github.com/mjc/sftp-s3-rs. this shook out a bunch of bugs in russh, which was fun.
- didn't like that there's no par2 implementation in rust so I built this https://github.com/mjc/par2rs (I'm too lazy to move to tape backup so it works pretty ok for dvd/bluray parity), unfinished but good enough for my use.
- same deal for 7zip in rust. https://github.com/mjc/r7z
- a medication tracker thing that uses claude/codex/copilot to scan the bottles and parse them as well as identify pills etc. works better than you'd think but I'm not planning on releasing it for a while.
fixed or fixing bugs in:
- exqlite (it should not crash anymore and should return busy a lot less often.)
- russh
- swift-nio-ssh (this might be why codex's remote can't connect to your ssh box) https://github.com/apple/swift-nio-ssh/pull/236
- NanoKVM (working on making the streaming for this a lot more fluid)
1. A personal dashboard that is gloriously incongruent, but solves almost every problem where I have to glance at my phone: my home and car battery charge levels, my failing github actions, planes overflying my house, medication tracking, Life360 integration so I can ensure my kids charge their phones, sports and finance tickers, birthdays, fuel prices, public transport, integration with my bathroom scale....the list goes on and on. It has ticker mode, card mode and alert mode and lets me add features via a Github and Claude API integration.
2. A shopping list app that allows me and my partner to coordinate on our shopping
3. A recipes app that includes AI scanning
4. A standalone home battery dashboard/app
5. A fuel prices app that is tailored for the closest fuel stations and is ad-free
6. A tool to draw classroom supervision maps for my partner (thrown away already, I didn't want adware/bloatware so I built it, she used it, then I threw it away)
7. A quiz website, cos the one I used to play on was overrun by ads.
8. A time tracker that I'll throw away at the end of the tax year
And more, and that's just what I did for making my life easier, there are other more "enterprisey" things I am working on. They're web apps that I add to my iphone desktop or run on otherwise junky old tablets or on TVs.
The point is that they do exactly what I want them to do instead of relying on downloaded apps that get me 80-90% of the way there, even if they'd be classified as "AI slop". I know enough about security and caching that they aren't full of holes and don't kill upstream, but I don't really care about the code, and it's literally easier for me to build something new than to go to Google or an app store to find software that's full of ads.
Over the past few days I have been making a spell checking TUI app. I used AI (meaning: free Gemini web interface) to discuss various aspects about the apps and debug compiler errors ang suggest useful rust crates for various problems.
Just a more helpful discord chat generally. It also gaslights you too!
A custom harness backed by dagger, gives diff, time travel, forking of both files and env. Building a harness is a good learning project. I'm now using other tools to see what they are like. (OpenCode is quite good out of the box)
Currently working on a markdown search and wiki backed by Typesense, also has good web search, fetch, crawl. This will power my personal knowledge base system as an important step towards more leverage and better outcomes.
A Javascript framework called places.js for creating interactive UIs using web components. It has support for cross component state management, backend data fetching, and web scraping protection. https://codeberg.org/createthirdplaces/places-js/src/branch/...
Built a meeting-intelligence pipeline that turns raw, error-prone transcripts into a structured, queryable knowledge base. Meetings get auto-transcribed by Krisp, whose speech recognition mangles the things I most need correct, like colleague names, customer names, internal product and architecture terms. I hand each transcript to Claude alongside a hand-built context document, and it works a fixed routine: read the context file, read the transcript, then reconcile every uncertain name or term against a master error table before drafting anything. Only the genuinely unresolvable handful surface as questions; everything else is corrected silently. Once I confirm those, it emits a cleanly formatted markdown summary in a manner I describe as a template: overview, topical notes, decisions, action items — and pushes the work items into Todoist so commitments don't get lost.
What makes it more than transcription cleanup is the back end and the feedback loop. Each summary hits Obsidian with YAML frontmatter and live Dataview queries, so open action items and meeting metadata behave like a database rather than static notes. In Cowork the whole accumulated Obsidian folder becomes fully queryable rather than merely searchable — instead of grep-ing for a keyword, I can ask questions that reason across months of meetings ("what were Todd's table-stakes asks, and has anything shipped against them"), with the model able to look across separate conversations. The other half is self-improvement: every clarification I resolve gets written back into the context document: its people directory, terminology glossary, and especially the ASR error table, so a garble I corrected once is corrected automatically from then on. Over time that one document has become a domain-tuned lens, and each meeting both draws on it and sharpens it, which is why the summaries keep getting tighter and need less of my intervention.
Beyond that: I use a Netatmo weather station which has a RESTful API (or sends to a cloud server that has one) - I pull that information (which I can see on the web and their apps) into my own VictoriaMetrics / Grafana set up on Kubernetes, via a Go app Claude built.
This app above was when I had my little aha moment: Netatmo's OAuth is slightly broken (issues with the different tokens and refresh). But I'd written a dog-ugly app which managed to work a while ago. Claude kept trying and strugglign to understand why its OAuth code wasn't working, and was asking me "are the credentials right?" etc. "Yup, I'm able to get data from my old app", then it said "If you have the source code to that app, I can figure out what's up", it looked, identified the issue, tried to work around it and then we "agreed" - "Hey, this should work this way, but it doesn't, and whether my old OAuth code should work or not, it does, so drop that in, and keep going". "Great, let's do that."
Not tools but my Quickshell config. Of course AI made a ton of mistakes so I cleaned it up a lot myself. But I was able to go from not having ever written a line of QML nor reading the docs to having a working top bar pretty quickly.
Came here looking for some examples, i tried some pcb routing, its horrendous. But you inspired me to maybe just run with it? Like design a alarm clock and see what abomination ‘it’ will come up with and just run with it/produce it anyway? Would be fun!
I've always used notepad++ with one single giant .txt file for taking notes with dividers separated for each day so I codified that practice into a desktop app
Just a super minimalist thing where each day is one .txt file with the newest one at the top and a lazy loading scrolling list with every note going back 5 years
Supports CTRL + F searching, backups, and a bunch of other QoL features/macros
Its kind of a revolution that with agentic coding everyone can have their own hyperspecific customized apps
I restore old non-functioning radios with nice design and good audio. I replace the old tech except the speakers with a raspberry add a mic array and package them with an easy to use ubuntu. whisper and the small gemma models made everything so much easier and private. I basically rewrite the whole backend with claude and created a nice first setup experience akin to other smartspeakers. It's amazing how good the sound is from these old speakers.
I schedule reminder calls to myself before some important appointments. It keeps calling me until I receive the message which it reads me (I set the message when scheduling the reminder call) and I have to say "message received" which marks the notification as delivered. (I use Twilio to place the call.)
I find a phone call is more likely to get through to me than a reminder or alarm, which I can ignore or forget; an ordinary reminder is not as interactive.
Claude built it all and although there's a script for it, I just set the reminders in an interactive Claude code session in the directory. (Like I'll open a claude code session there and say "using the script in this directory, call me tomorrow at 7 a.m. with the message 'dr's appointment'."
At work, I've created a few convenience scripts in bash and Python - the second of which I am not fluent with. So, I used anonymized LLM access to create boilerplate/simple scripts with a bit of argparse and NumPy, which I then adapted to do what I actually wanted.
Would have made them without UI with a bit more elbow grease invested in web-searching for some examples, maybe even a StackOverflow question.
Generally, I'm not a fan of LLMs and their social effects.
It's a user daemon that runs on my machine and exposes a unix socket, and then a bunch of hooks in claude, zsh, vim, etc, that report directory and commands I've run and all that, pipes it to claude Haiku for summary, and then stores context in sqlite. It also exposes that data as MCP so I can use claude to say "hey what was I doing yesterday," or any arbitrary time range.
I find that in the age of using AI agents, "Wtf was I working on yesterday" is an even harder thing to remember for me, so this helps me kind of track everything with a database that a) has AI summaries already and b) can be accessed by AI as well as a CLI.
The "expose it as MCP so AI can query it" pattern is one I keep seeing work really well in practice. We did something similar for business metrics in Databox - the interesting design question was whether to expose raw data or pre-aggregated metrics. We ended up with metrics (with dimension breakdowns) because agents hallucinate less when they don't have to decide how to aggregate themselves. Curious if you ran into anything similar with the sqlite summaries - do you find the AI-generated summaries more useful than raw command history?
Because I work across a dozen repos any given week, and I have my own side projects I play with as well. It blurs together. Global git hooks are another data source for the project, which I'd forgotten to add, so that data is in there too.
I've vibe coded multiple helpful apps and websites for recording data. But longer term, I'm building with its help an internal research system to organize, search, compare, analyze, and esp reuse all the large amounts of data my firm produces, with the public materials without constantly starting over in separate ChatGPT or Claude conversations.
Similar to you, the things I have truly vibe-coded (having looked at <5% of the code) are largely data focused. Data labeling, organization, etc. These applications are extremely janky, I'd never ship them to users. The UI is mediocre at best. The functionality hardly better. But the point is to get data out of them. The code is a means to an end and not a product in itself. Building a custom dataset builder in just a couple hours of work is really powerful.
magpie - extracts book recommendations from reddit threads. I had a bunch of saved threads from 'books' and 'suggestmeabook' and 'printsf' etc., and I realized I could pull them down and do a semantic search.
Vibed a ton of small projects, mostly for my own usage, to understand what agents could do, or just to satisfy my curiosity. Most are rough around the edges, and a bunch quickly became obsolete as agents got better and I stopped using them.
https://modelarchviz.tsilva.eu/ — View model architecture diagrams, codebase, and paper side by side. Includes a chat bot that can see selected content and select content itself. WIP.
https://embeddingviz.tsilva.eu/ — Visualize model embeddings and/or layer activations for different content with PCA, UMAP, or t-SNE. WIP and poorly tested.
https://github.com/tsilva/dlab — Workbench for deep learning experiments. Open the repo in Codex, pick a training target, e.g. maximize validation accuracy on CIFAR-10, and ask it to start a research track. Then collaborate on training runs, evaluations, sweeps, and iteration toward the goal.
https://github.com/tsilva/gymrec — Record and replay gameplay from Gymnasium environments as Hugging Face datasets. Supports stable-retro environments such as NES, SNES, Genesis, etc.
https://github.com/tsilva/agentpong — Run coding agents in VSCode/Cursor terminals, one per Aerospace window. Agents trigger desktop notifications when done; clicking a notification sends you to the correct desktop. I haven’t used this in a while, so I’m not sure it still works.
https://github.com/tsilva/agentbox — Run agent CLIs inside Docker sandboxes. Project-level config defines sandbox access. I stopped using this once agents got better built-in approval mechanisms.
https://github.com/ferrislucas/Circus-Chief
Circus Chief is a tool for managing coding agent sessions from a browser. It's specifically optimized for small screens. It supports Claude Code, OpenAI Codex, and Google Gemini CLI agents.
Features
Agents can operate Circus Chief itself. Agents can spawn sessions, schedule sessions, interact with the Kanban board — anything you can do in the UI, an agent can also do.
Schedule work ahead of time.
Automatically reschedule when you hit usage limits.
Configurable, chainable prompt templates.
User-defined commands - agents can run them and see the results, and so can you in the UI. Handy for local CI or routine workflow steps.
Worktree-per-session isolation, or elect to work on a specific branch.
Shared canvas - this is a place for shared artifacts that don't belong in the code. Useful for iterating on planning documents.
Bring your own provider. Use subscription auth for Anthropic, OpenAI, or Google, or point sessions at third-party providers with Anthropic- or OpenAI-compatible endpoints.
If you want to try it:
Lots of small dashboards/log investigations deployed to private Github Pages for $WORK. This has been a great way to share insights.
I'm currently working on a tool to control my tmux sessions from my phone. Specifically all the codex panes so I can remote control my running sessions from my phone while I'm away from my desk while also keeping them in sync with tmux for when I'm back at my computer. On this one I'm using Codex as a helper, it's not allowed to write any code so it's going along a lot slower :)
- Habit Tracked which allow to track not only habits (done, not done) but also numbers and time of the day - https://habitpocket.io/
- Todo list, my own interpretation of Todoist and Things 3 - https://taskpocket.io/
Both of the apps above was built using AI especially for the frontend. Both backend at some point of time I almost completely refactored/rewrote myself
Speaking of private apps:
- I made a small app which allows me to track business expenses and categorize them Built with AI and using AI.
- Android weather app that tells you what to wear: https://play.google.com/store/apps/details?id=app.clothescas...
- Android launcher with fuzzy find: https://play.google.com/store/apps/details?id=app.typelaunch...
- Webapp to view your ancestors on a world map: https://gedmap.com
- vcs wrapper around git, hg, and jj: https://github.com/mikelward/vcs
- Migrate my bash and zsh configs to fish and nushell: https://github.com/mikelward/conf
Uses local AI models and I was able to snag this great domain name.
https://finalfinalreallyfinaluntitleddocumentv3.com/
But hands down the most useful thing I've made is HutchDB, which is a MCP service that you can call from any AI chat or Agent setup to store data for you.
Literally from your AI you just say "save that to Hutch" and then it figures out:
- The schema + fields - Builds nice webviews (Kanban, Timeline, Grid, Calendar) - Lets you share the output with people
So people use it for all kinds of things like time tracking ("every hour save a summary of my activities to Hutch"), for Agent to Human handoff ("Every day check social media for mentions of my company and save them to Hutch").
I use it for things like recording all of our marketing activities and then having my AI compare those to signups for rough attribution, etc.
Dead useful and at https://hutchdb.com
https://www.noisetable.xyz/ - a collection of chance-based audio 'channels' in a VCR inspired interface
https://concrete95.net/ - a musique concrète web app that's made to look like windows 95. Pulls random audio from freesound.org and loops a small section, you can also layer synth pads or melodic synth loops. I'm often able to get some really pleasant background ambience.
https://windso.me - a sample-based step sequencer that doesn't let you choose the sample that's loaded, kinda fun, still needs a lot of work!
- sandvault https://www.codeofhonor.com/projects/sandvault/ runs agents in a separate macOS user account, hardened with sandbox-exec. It also supports headless browsing and iOS Simulator from inside the sandbox for testing web and iOS apps.
- clodpod https://www.codeofhonor.com/projects/clodpod/ agents run inside a macOS VM.
- git-multi-hook https://www.codeofhonor.com/projects/git-multi-hook/ git only allows one script per hook event; this is a dispatcher that discovers and runs every script in a hooks dir, in parallel, for both global and repo hooks.
- TubeGate https://www.codeofhonor.com/projects/tubegate/ Chrome extension to block YouTube videos based on keywords (like “sponsored”).
- push10k https://www.codeofhonor.com/projects/push10k/ iOS app to track my progress toward 10,000 push-ups.
My blog is AI-coded: Zola static site, Sveltia CMS, Cloudflare Pages/Workers, with GitHub Actions handling builds and syndication. https://www.codeofhonor.com
This place is incredible.
Guild Wars 2 and most other games are pure slop.
Also Guild Wars 1 has been receiving new content updates this year.
>>Some content-management software (CMS), like WordPress, requires using the same presentation layer that the CMS uses
Headless wordpress has been a thing for quite a while and it’s trivial for a use case like this
Newest Git supports hook events.
At the same time, I moved from Chrome back to Firefox, and Gemini was great at finding equivalents for my most-used extensions -- and, when none existed, to write my own. It's also been really useful for customization/"ricing".
More recently, I got into Quod Libet as a primary music library manager, and both Gemini and Claude have been fantastic at helping me build custom plugins that make it do exactly what I want it to. Scripts to automagically download tracks with metadata and synced lyrics, a lyrics sidebar that highlights lines as you listen and lets you click to jump to a specific line, a bookmark button that lets you mark your favorite section of a track for easy browsing later. Next chance I get is something that enables lyrics search across the entire library (a feature I was already able to build for the Stremio desktop player -- it's so cool to be able to search for a line in a TV episode or movie and jump straight to it).
For me that’s killer feature, even if we don’t achieve AGI at least we got good enough „something that will google it for me”.
It is great both ways as an expert in my niche I don’t have to waste time on reading through entry level fluff. As non expert in other fields I don’t get to be scolded for asking entry level questions so RTFM and LMGTFY will sink in history fortunately.
I can't help thinking that this is a combination of Google sucking more and more and various problems of daily getting so hard need something like a script to solve them. (Recent challenge - "what affordable campgrounds are near the Pacific Create trail and open now").
Similar to another tool called process-compose but with a CLI that mimics the docker one more closely and doesn't start a TUI by default.
Useful if your team's on nix and doesn't need containers.
I made a scraper that searches through all the news and finds anything 'war' related. It then summarizes it and give me that new in the Star Wars crawl with music.
Code: https://github.com/kordless/force-news
Wait, it would be easier for me to clone the whole thing and change one feature... What strange times we live in.
Almost everything I build will run locally or on Google Cloud in serverless mode.
The README on the repo for the app is has been updated with instructions for you. You will need Docker.
Plus after each run you get screen recordings with console logs, network requests, HARs, and Playwright traces so you can inspect exactly what the agent did :)
https://github.com/wizenheimer/canary
P.S. I attempted to do a Show HN but got flagged for some reason
From a quick look at your profile, the majority of your submissions have been Show HNs. HN only allows some fraction of your submissions to be Show HNs (imagine if the front page was nothing but), so eventually they will just be auto-flagged.
I am the only user. Sometimes it's the process that matters, and exercising your brain is important too. I get that there is a lot of existential dread around AI taking our jerbs, and excluding humans from the process of creative work, but... you can still just write code, just for the personal satisfaction.
https://github.com/armyofevilrobots/bap-egui
- https://github.com/exlee/pikchr_pl - DiagramIDE (diagram amiga-style workbench where you can script Pikchr diagrams using TCL, Prolog, Pikchr or - recently mruby). Note: you need to navigate to actual crate for description. There are binaries built in case someone wants to try it.
- https://svg.axk.sh - semi manual SVG fitter so that I can easier vectorize AI-generated pseudovector images (who doesn't like 30kb SVG versus 1.8mb PNG?!)
- https://github.com/exlee/rik - this one makes me laugh every time I use it - it's an AI harness with text editor as an UI (i.e. it reacts to comment strings) - I gave it personality so it makes wacky comments but other than that it's very constrained agent (limited edition ;))
These are not vibe coded but AI made it much easier to slide through major friction points (e.g. for SVG fitter I really didn't want to reinvent fitting algorithms)
This whole thread is turning HN into my favorite app store. Good stuff.
A game engine / MVP game
A tool to replay shell commands during presentations
A tool to generate ttrpg book pdfs from obsidian markdown
A tool to generate confluence pages from markdown
So basically a way to just go on an hour long walk with myself, spit everything from the top of my dome stream of consciousness style, and then have Claude structure whatever I said.
It's nice to have something that structures my thoughts by just thinking out loud.
I vibecoded it (it's approaching 20K lines including tests). It works quite well but there are some bugs, so will have to do some actual engineering. But the UX is working quite well.
But I have to ask: why not just advanced voice mode in ChatGPT or Claude?
To make matters worse, they've recently gotten rid of the timer, so you have NFI how long you've been speaking for.
I use it for therapy-based stream-of-consciousnesses + venting and then have my project set up to understand the schema therapy work I've done with my psychologist and give me insights / draw threads between things from my past and now, and losing 10 whole minutes of talking and processing is SO FRUSTRATING!
To be fair, vibecoding this memo app in Swift didn’t take too long. There were some tricks to it, using xcodegen helped a lot so that I don’t need to use the Xcode project.
It’s fun to see Swift code. I used to do some Objective-C back in the day.
[1] another thing I made. It’s a sequel to the Alice in Wonderland stories. It’s also a SQL course. I vibe engineered it, meaning I looked at the code and used AI-assisted development.
Except for the story though that’s almost all fully me. LLMs aren’t great storytellers. The same is true for the lesson scaffolding, that’s almost only me.
Turned out super helpful to keep track of everything in one place (shared expenses, ask / demand unplanned changes in custodies, kids activities (school events, birthday parties, sport competition and so on)
Its helpful even for only one parent.
https://custody.homes
I dropped any plan to make money from this so its free
(Edit: forgotten in first edition) A cookbook to store the recipes my family likes to cook so I can eventually break up with Pinterest: https://github.com/vtbassmatt/Cookbook
A data extraction pipeline and search engine for a new card game called Mood Swings: https://moodswingsdata.github.io and https://moodswingsdata.github.io/feelings.
An app to let my friends and me build a Magic: the Gathering cube iteratively together: https://github.com/vtbassmatt/popcorn-cube
A custom wiki engine for a family of podcasts I enjoy: https://github.com/vtbassmatt/totalus-wikium
A systemd log viewer for the web: https://github.com/vtbassmatt/djournal
- a daemon to properly set the LEDs on my Lincstation N2 NAS (even early Claude was great at reverse-engineering): https://blog.majid.info/lincstation/
- most consequential, Thufir, a task manager to replace Cultured Code's Things as I am leaving massa Tim Apple's plantation: https://github.com/fazalmajid/Thufir Still very janky, but usable enough for my purposes
- most recently, a TRML dashboard for my Seeed Studio reTerminal E1002 smart e-ink display, and I also had to have Claude fix the firmware for full 6-color input: https://blog.majid.info/e-ink/
With Tailscale, you can basically point a domain to the FQDN of a machine you’re sharing with people and the domain will simply work for them (and only them)[1]. But for it to work without them having to know or specify the specific application port, you have to grant them access to 443 (and 80) in your Tailscale ACL for that specific host.
So yeah, now immich.familydomain.com works without family members having to remember the specific port. BUT, serveradmin.familydomain.com (another app on the same host) will ALSO be accessible to them (from a networking POV). We opened port 443/80 for that host after all.
I took a few hours with Claude back in January (?) and we wrote a tiny Go authorization gateway which basically consults both Tailscale’s public API and Tailscale’s `localapi` and returns the appropriate response to Caddy based on the requesting user’s actual allowed ports.
So now I can share different apps (subdomains) with different people without forfeiting access controls, all driven by Tailscale’s policy file.
(I hope I didn’t mess up the crux of it, pretty late here)
Edit: why not (something like) Authentik? Quoting from the draft:
[1]: https://www.youtube.com/watch?v=Vt4PDUXB_fgIf tomorrow I decide not to share with anyone, I don’t want to have to reconfigure stuff. I simply edit the Tailscale policy file, and (maybe) spin down my server.
I used OpenClaw to make a health and wellness coach agent that tracks calories and alcoholic drinks and logs it to a personal dashboard. I send it photos of my meals, and it will estimate the calories and log it. It will also help me make meal decisions and give me words of encouragement.
I used this HAM dashboard git repo to create a bespoke dashboard of different video and weather feeds from my area: https://github.com/VA3HDL/hamdashboard
I also, in the same rabbit hole, created a radio reference guide for the Sonoma County area: https://mybbor.com/petaluma-sonoma-ham-radio.html
I've spun up probably close to a dozen one-off or small websites for various little interests or projects. One of my favorites is a short domain file uploader that I can quickly host Markdown and HTML files to share with family, friends, and colleagues. It's using Caddy and running on a DO-VPS. I open sourced the code here: https://github.com/RobbyMcCullough/honeydrop
And instead of turn-based, you have to hike to/visit a physical location, fill in a code you found there to get bombs for the game. Or do a quiz and get the answers right.
Great fun so far!
Could I have done this myself? Of course. Would I have tho? Prob not.
This kind of simple, hyper-specific bespoke utility is the perfect thing for vibe coding IMO.
I've always found it annoying that browsers autoplay animated images, and there still isn't a built-in way to control that behavior.
The extension shows the first frame and adds a play/stop button directly on the image.
What started as a personal utility ended up being published on the Firefox and Chrome extension stores. It's still a small project, but it solved a problem I had every day.
Firefox: https://addons.mozilla.org/en-US/firefox/addon/gif-control/
Chrome: https://chromewebstore.google.com/detail/gif-control/nhoihin...
Bug reports and feedback are very welcome.
Disclaimer: I previously posted a Show HN about it, but it didn't get traction [1].
[1] https://news.ycombinator.com/item?id=48208916
> I previously posted a Show HN about it, but it didn't get traction
FYI your linked submission is marked as dead. Not sure if that's a problem with your account or not. You should email the HN mods.
https://tajd.github.io/cofferdam/ - is an experimental compile time time checker that can be run for typescript/rust to be able to impose architecture, design principles that can be forgotten by agents. The goal is to be able to make it possible for an agent to be reminded about design conventions where it forgets, so it reduces the amount of context that needs to be used initially to define those sorts of conventions.
Hister is a full text indexer for websites and local files which automatically saves all the visited pages rendered by your browser. It provides a flexible web (and terminal) search interface with offline result previews & detailed query language to explore collected content or quickly fall back to traditional search engines.
It can provide a privacy-respecting search experience for serving "recall" type searches where users retrieve previously visited content, but falls short in "discovery" type searches (yet).
Much kudos. I hope more people discover how powerful even a local search index of previously visited content can be. And I hope more people can build large indexes as well, so we're not just relying on Brave & Mojeek & Marginalia (and EUSP) to rescue us for the fallback discovery searches.
Also used AI to design an online store (I'm not a front end dev). It's amazing to see my wife (non-technical background) tweak the web-site using claude code.
End result: an online store where we sell jewelry pieces that actually are lambda-diagrams (Tromp diagrams) that compute Graham's number, or of the Y-combinator (well, technically it's just a fixed point operator, one beta reduction away from the Y-combinator. But Y-combinator was not aesthetically pleasing from a jewelry point of view)
store: Built a tool to help design cs/science inspired jewelry in CAD. I wrote a DSL to describe the jewelry, and had an LLM write the interpreter to generate a CAD file using cadquery (note to self: LLM suck at 3d reasoning). I would not have had the time to do this without AI.
Also used AI to design an online store (I'm not a front end dev). It's amazing to see my wife (non-technical background) tweak the web-site using claude code.
End result: an online store where we sell jewelry pieces that actually are lambda-diagrams (Tromp diagrams) that compute Graham's number, or of the Y-combinator (well, technically it's just a fixed point operator, one beta reduction away from the Y-combinator. But Y-combinator was not aesthetically pleasing from a jewelry point of view)
store: https://studio-galois.com
A WebGl program that takes input like X123 Y123 Z123 via WebSerial every 100 ms and builds an object out of the resulting path. Required some performance optimizations (just had to tell it what to do). Also asked it to make the corners nicer and it did. (To be fair, I'd already asked a lesser model and put some things in the prompt to nudge it the right way.)
Various OpenSCAD models. E.g., remote control holder with 5 slots, staggered heights, slight slant because it looks cool, and the slots all have different depths. One shot. It implemented the slant/tilt using a shearing matrix. 100 points.
one trick on 3d reasoning: get it to draw all the different orientations, and you pick which one to use
it save a lot of time vs trying to tell it to rotate around Y and it actually rotates around X
What I often ended up doing is asking it to draw 3 labeled arrows X,Y,Z. So I could tell it to orient along the XY labeled arrows (which are in reality YZ, but whatever).
We have not really started advertising, but my wife is (very) often complimented on the jewelry when she wears it and that has led to a few sales.
https://github.com/nibrahim/glocker
A self hosted web archiving tool with support for extendible processing pipelines (eg. extract article -> translate -> summarize -> generate tags, download video -> split audio track -> transcribe -> summarize), which led me to make a managed chromium browser with extensions and warc support for archiving, and a RSS feed synthesizer (take random article listing page that doesn't have RSS and generate a feed for it) so that I can plug it into my archiver. An active learning loop for a model to clean up articles by removing junk like native ads and sponsored blocks.
A tabbed terminal with project management features like launching the database, app server, and claude code in different tabs with one click, and split browser/terminal panes (eg. opening a browser automatically at the correct URL when the terminal reads http://localhost:4000/).
A modular MCP server with a MCP proxy and OAuth2 dcr so that I can easily add new random ideas for MCP servers in a few minutes with Claude and deploy them such that it's available to Claude by refreshing the tool list.
A small tool to render Claude conversations so that I can link to them from my obsidian vault with something like convo://claude-code/-home-jfim-projects-foo/<guide>
And overall just deploying docker containers for my self hosted setup
Most of it is on GitHub, in various states of readiness.
When you plug it in, the device is recognised, but press any button and it attempts to start the pairing process. Then using evtest nothing is coming through.
That^ was pretty much my prompt too, and 10 minutes later I have a working driver with systemd unit so it works through restarts. Amazing stuff!
Besides, I made a lot of automation scripts (mostly using Ruby) that run on my raspberry pi to fetch/parse/crunch things and notify me on my Android phone through a self-hosted https://ntfy.sh server.
- Automated backups of steam saves for when my kids wanted to play the same game on my account and saves get out of sync with steam cloud. The kind of thing I wouldn’t usually bother with myself.
- automated script that reruns failed GitHub actions in repositories with flakey tests ‘cause why bother fixing them? It also auto catches up branches with the main branch for the repo.
- a YNAB extension for pi (agent harness) to help with entering purchases that need to be split across categories. This is also in the telegram bot so I can use voice-to-text to explain a purchase.
- I already had some python scripts to generate pdfs of Magic: The Gathering cards for printing proxies. I had an llm extend it to make some dividers with the set names and symbols on them. Makes organizing them a lot easier in the big card boxes.
A knowledge base for my research area, with tools for paper ingestion and search.
An md file to html presentation tool, there are several but this one helps me.
A review tool that splits a PR or branch intelligently into modules, and does per module reviews and global reviews for different aspects, and then summarizing that into a report. Can be used with multiple different harnesses. Written as a Python project, but build-time assembled into a single-file Python script with uv run --script shbang line.
https://github.com/ankitson/webby
CF pages still required too much confusing clicking around on a webpage for me. This way I can just point any little report or app at a directory and done.
There’s others that are more server shaped and tightly coupled - a pipeline to pull in all my data like Garmin, Twitter bookmarks, messages into a Postgres DB. Kind of a personal data warehouse that i can use with apps/automations, like alerting me if my sleep schedule is drifting, and a custom web interface to my Garmin data
Later I find the technique I developed to instantly load arbitrary large CSV files can be generalized to work with any format, with an incremental parser combinator. This means the tool can read from not only files, but any stream-oriented sources like a pipe.
I did used AI in development but it didn't speed up the process very much, as I found a lot of time was spent on the deign of an intuitive and consistent UX. The project is still not very production ready, but in case anyone is interested: https://github.com/Verticalysis/Hitomi
I also built https://plugins.audio - a Dribbble-esque showcase of audio plugins.
It's been a dream to be able to work in the audio software space. I've been a musician, designer, coder all my life, and had a few moonlights with building audio plugins over the years. But now with AI at hand, I can use it to fill that missing technical requirement while still retaining my domain expertise (music/design).
I also built a kanban board with agent integration and context management, with a vs code extension to go with it (also helps with git worktrees too): https://www.agentkanban.io
There is AS Notes - an Obsidian / Logseq / Roam alternative for use a s a VS code extension (is designed for use behind corporate firewalls, git friendly): https://www.asnotes.io
Also NumeroMoney: https://www.numeromoney.com - For personal finance spending analysis and budgeting.
AI has been a great 'exoskeleton' for me. I fortunately had some good infrastructure and solid application base templates from before AI 'got gud' and so building on these has been the best of both worlds - a solid base and improved speed of development.
I also wrote an honeypot that emulate an Ollama instance: beside the attackers, it's funny how many people are looking for free inference. Somebody from Brazil try to use my honeypot to write to chapters of a book about traditional magic rituals. My next step is to extract the data collected by this tool to extract IoC and malicious prompts and share them with the community.
In the same scope I wrote also an Ollama scanner: it fetch from Shodan the open Ollama instances, verify that they are reachable and check if they are real sending a dummy query.
I probably would not have bothered to allocate time to this pre ai, the juice wasnt really worth the squeeze. But I approached it with an initial amusing naivete about it being 'super simple'. As is almost always the case with software theres a reasonable amount of hidden complexity. But I have been using it as a sort of learning proving ground for how to work with agentic development. For example I got to a point where claude wouldnt implement properly and would argue with me about changes because it would read the current/old docs in the repo and get confused about reality. So right now I'm experimenting with 'canonical specs' that can only be changed modally with gates and a defined cascade from canon, to code, to docs in that order. Otherwise you end up in a weird thrash about the docs and the code disagree and which one will the agent decide to change for consistency?
Anyway, its been interesting and its v0.6 and at a point where Ive not hit a sharp edge dogfooding for a while and some beta testers would be valuable. Right now you have to manually wire it into your stack, once some others have kicked the tyres hard enough I will make some pr's to the popular tools to consider integrating it.
https://gitlab.com/grepular/foxcage - Runs Firefox inside podman to isolate it from the host. Has some interesting features that I wanted and nothing else gave me.
https://gitlab.com/grepular/claude-sandbox - Yet another Claude sandbox. Runs it inside podman again. Has a pretty powerful proxy system for securing your credentials.
Currently working on a tool for sanitising email. Will be blogging it up at https://www.grepular.com/blog/ when it's ready for others to use. Does things like applying policies to html/svg/calendar/vcard parts to whitelist or blacklist tags/attributes/css/url schemas, clean URLs, fetch remote content at delivery time and attaching to the email to prevent tracking, pgp and smime auto encryption/decryption and a million other features.
Various MCPs for above.
A "remote claude code server", that gives project level overview and lets me run projects / develop on my home server rather than locally through my laptop. Supports ssh as well as a web UI (projects in a list, shell rendered using https://xtermjs.org/, with a tile overview when working on multiple projects to watch for turn ends.
Similar to above, I have a local version that auto launches a new project scoped podman container, passess through the work directory, installs CC/Codex/Grok into it and passess through the auth / config for each agent. Then dumps you into that shell with aliases to map each agent to that agent with a few special env flags to disable permission prompts, so claude = `claude --dangerously-bypass-permissions`.
An extensive MCP for Obsidian that gives agents access to use a lot of the more advanced Obsidian functionality, such as suggesting and installing plugins / configs / etc.
It's helped some of my daily productivity, but I still prefer to get my hands dirty with code most of the time rather than full prompt it.
I'm thinking about creating something pretty similar, I want a digital housekeeper that keeps an eye on what is happening in my house and notifies me about dead/unreliable devices, fix broken automations, suggest new automations based on sensor data, etc...
I've setup the unofficial home assistant MCP already, but LLMs seems to struggle a bit to use it properly and I haven't looked into it yet to understand what is happening.
https://github.com/pkhamre/opencode-docker
Also made this minimalist carousel generator after seeing a carousel I really liked at LinkedIn.
Project: https://its.pkhamre.com/p/carouselify/
Source: https://github.com/pkhamre/carouselify
Normal browsers have built-in dev tools - this has built-in design tools. so I can visit my app, open up a surrounding canvas, pull fragments into the canvas, do some design-ish stuff, and merge it back into code. All in the same UI. It was cool enough that I'm going to release it, but for now it's very useful for myself.
https://matry.design/
It feels like you have parallel conversation with an agent, except it understands all the surface. And it's granular, surgical and precise.
Usage: I'm writing a game design document of 70 pages with it, working surgically without having to worry about what page number the edit was.
It really solved my main bottleneck which is telling the AI what to do in a complete and comprehensive context.
Readme is trash but your agent will understand what to do
https://github.com/WiseDragonAI/TheBlueprintTool
- CleanMyMac alternative based on opensource tools - https://github.com/p-raj/open-cleanmymac
- Standup meetings to comic generation - https://github.com/p-raj/standup-to-comics
- Configurable Pomodoro - https://github.com/p-raj/open-tomato
A few more closed source ones that aren't any close to be in a working state.
I'm probably the only one who'll ever use it but I like it.
And yes, there are lots of cows. The longer you stay on distracting sites, the more cows appear.
The idea is to make distracting sites less appealing, without using a black-and-white site blocker, which you can easily disable.
Chrome: https://chromewebstore.google.com/detail/mooblock/eanbagjehd...
Firefox: https://addons.mozilla.org/en-US/firefox/addon/mooblock/
Open Camera Control https://github.com/jcubic/open-camera-control - that allows me to control the settings of my DSLR while I'm recording myself.
Horavox - https://github.com/jcubic/horavox - A speaking clock
Mutimon - https://github.com/jcubic/mutimon - a config driven web scraper (found this post from the email sent by this tool).
ASCII-Globe - https://github.com/jcubic/ascii-globe - JavaScript library that renders a rotating earth or any map. Can be used to add animation to your website.
It creates its own copy of your workdir for the agent to play in, and then you pull changes out ala git diffs or commits.
It's a MASSIVE time saver, and I use it as my daily driver.
https://github.com/kstenerud/yoloai
Not exactly a tool, but I also made pelohard.com which ranks the most recent Peloton classes by difficulty. Updated twice daily.
Most of the tools I write now are bridges to various SaaS products that have APIs but no CLIs.
I also got it to write a mysql DB importer I could use instead of "mysql", since I often import large DBs locally. I wanted progress, estimated completion time, and I wanted it to prevent my machine from going to sleep during the import.
https://github.com/hale-lang/hale
* plugin for Logic Pro to simulate how a mix will sound on my macbook and phone (I captured real impulse responses for that, sounds very close)
* an app for spaced repetition for guitar video lessons / their parts (no idea why platforms like truefire don’t have this feature)
* workout planning/tracking app
* an app to create impulse responses for acoustic guitar, to make it sound good live
At first glance, it's just a nice version-controllable, parseable DSL. But I also made succinct prompts with the grammar that lets LLMs produce and reason about Satsuma, a language server, CLI tools for the AI tools to use to navigate specs token-efficiently, reason about lineage, pretty viz in vscode plugins/syntax highlighting, some agent skills etc. There are metadata conventions for succinctly representing a lot of quirky formats and capturing common analytics conventions (scd2/Kimball/datavault/etc c.)
Yes, yes, I may have gotten carried away.
But I'm finding it really useful as a specification tool in projects for reverse engineering mappings from code/workflows, generating new code (dbt, Spark etc.)
This definitely isn't something I would've had the bandwidth to push this far before AI!
https://github.com/egorelik93/Doom-Emacs-Config/blob/master/...
The most complete tool being this unit testing support library for Apple IIgs assembly language development.
https://www.npmjs.com/package/iigs-unit
- An analysis pipeline that takes data from D1 (storing user searches), Posthog and Clicky (analytics) to periodically generate user journey reports.
- halalcodecheck.com - a database of emulsifier codes and food ingredients. Building this with the goal to help consumer verify and find alterantives of food ingredients across US, EU / UK and Asia.
"Hi [FirstName],
I came across [Company] and was impressed by the work you're doing — especially [specific detail]."
It's a terrible cold email, I get so many of these every week... There's better ways to get someone's attention
Having said this, that's just a demo template and users can tweak and configure the template as they deem fit.
https://emily.infiniwa.com/
I also build a ton of household glue stuff; I was never really passionate enough about the whole "homeserver" thing to spend the effort in going beyond basic video recording for my security system, but now I have all of my local-only home automation stuff wired together, mostly into HomeKit, and have been able to ditch a ton of cloud services.
Most recently, though, is a basic python CLI/Flask app that makes it point-click to manipulate the route table and dnsmasq settings on a raspberry pi. eth0 on the pi goes to LAN and put a USB ethernet attached to a switch and you now have a pretty powerful IoT/Untrusted device inspection environment.
One click to change the DHCP settings offered to the device, run a pcap, allow/deny the device access to certain services or NAT them elsewhere as needed. Straight forward DNS adjustments that can be applied per device and now that MITM-Proxy has a python API... it's pretty straight forward to also peep inside of SSL protected things.
Cloak windows from screen capture! Perfect for keeping things private while screen sharing over Teams, etc.! All other tools in the Store would trigger virus warnings when I tried to install them. Some were positioned to ask for money for basic features. My app is both free with more features and a more intuitive UI and set of keyboard shortcuts!
- usagi, for tracking Claude usage in the macOS taskbar. Lots of these around, but I wanted one that wasn't buggy and constantly adding features: https://github.com/duggan/usagi
- RockstarNinja, for sharing Claude plans and sessions. Since it's a bit of a data hog signups are limited, but me and my cofounder use it all the time: https://rockstar.ninja
- TweetEmbed, for browsing an offline Twitter archive and getting self-contained HTML copy/paste-able cards: https://duggan.github.io/tweetembed/
- bewitch, a terminal based metrics collector and data visualization system: https://bewitch.dev/
Each one of these has been a great way to really push on automating build processes, see what Claude can do for automatic documentation (screen captures, etc), and trying to give a distinct visual identity. In my next project, I'm trying to de-LLMify the prose it generates by using my own blog posts and aggressively pruning and integrating into the prompt.
The biggest change has just been unlearning "cost". Stuff I've subconsciously shied away from since intuition built up over a career has given me a sense of how long things take that just isn't true anymore. Still learning!
Can you explain the study thing in more detail? Or give an example of how you use it?
Then a little calorie-radar-alert app that motivates you to stay on diet before going into a shop: https://apps.apple.com/us/app/stay-lean-calorie-navigation/i...
And to test the hypothesis if the motivation actually works with "fast flashing images" I built this experiment: https://fitfatflix.fit/
When openClaw came out I build a game like map around it to orchestrate tasks: https://memefields.com/
And not published, I build a little recommender app of activities I can do only today or only this time of the year where I currently am, so I don't miss good things.
And some other stuff I'm working on right now :)
Sometimes I daydream that the end goal demo would be to set one up in a VM and let a sort of ransomware lose in it, and see if the agent can identify what’s going on and react in time.
But for now I’m fighting Apple notarization to enable local notifications on macOS.
[0]: https://github.com/lightless-labs/descartes
Of course, it was great to have exactly the features I want, and I enjoyed learning some new things.
opens terminal
(made me smile)
- https://github.com/artemave/artwall rotates my desktop wallpaper through random paintings (Linux, SwayWM)
- I get my TILs through https://t.me/daily_bite_sized_fun_fact
- https://t.me/tolmach_forward_bot helps me practice French reading
- https://mini-meet.artem.rocks/ was an attempt to circumvent RKN (russians) blocking video calls; not a complete success, but works for some people
- counted the number of dudes in Big Lebowski with https://github.com/artemave/super_video_grep (https://www.youtube.com/watch?v=V7EUtIbOd1w)
Now I want to add some sort of recommendation engine on top, to let me discover stuff I might like (I'm not into superheroes anymore).
Dont have the code up for sharing but I documented xeil (along with a few other tools) in my blog: https://paul.mou.dev/posts/2026-04-28-software-for-one/#xeil
Instant windows switcher with custom shortcuts and instant "opt+tab" and trackpad switching. Simple does exactly what I need it to do (just bypasses the slow window switching that is annoying), with no additional features or bloat.
I've also done a TUI that combines my messages from WhatsApp + messenger + discord which is pretty handy at work.
https://www.reversecam.com/
The front camera apps always show the mirror preview. Most of us hate photos others take of us, but love our selfies. It's because we groom ourselves and find our perfect angles using the mirror preview. So it's jarring when we see the photos of us taken by others. I always wanted a camera app that showed the non-mirror preview. Surprisingly most camera apps don't have this option. So I created this app to scratch my own itch:)
Note: Photos and videos are only stored in the browser. No data is sent to any servers. You can also install this as an app from your browser since it's a progressive web app.
Ruby on Rails: A volunteer 'jobs' board for OSS projects (ironically): https://ossvolunteers.com
JavaScript (client side+Cloudflare Worker): A map showing stop-level usage of OneBusAway across the Puget Sound region, updated daily: https://opentransitsoftwarefoundation.org/onebusaway/visuali...
Swift/iOS: An app that helps me track how much time I spend in the five heart rate zones, so that I can better focus my workouts. I'm working on version 2 right now, which will take advantage of some new features in iOS 27 and has an all new UI. https://www.zone2.app
Everything was built with AI as part of a 12-in-12 challenge I'm doing, where I'm building 12 products in 12 months. https://twelve.zamith.pt/
Quietbox (https://quietbox.zamith.pt/) - Clear spammy emails from your inbox
https://gregryork.github.io/EpicRpgBattleMap/
Or were you only asking about digital tools? I haven't really made any of those. Making physical tools feels much more satisfying these days.
I made that realization last year and since then it's just been random project after random project each one requiring me to discover a new tool or method to do some aspect of the project "properly". I'll never be a plumber or electrician professionally but it's so rewarding to start from zero and learn something new that is tangible in the real world. That's the one AI use case I've walked away from feeling like I actually learned something.
Low fire clay fires at 1060°C+ and high fire clay at 1222°C+.
[Corrected to both be Centigrade]
Im slowly trying to extricate myself by cooking more from home only from local farmers and what I can grow from home (so far only one cucumber). After all, can you really build everything else if your own body molecules are being replaced by low quality things made by others?
I'll get around to 100% at some point before I die or I wont care anymore since i'll be dead: one of those outcomes is inevitable.
In-browser 4x image upscaling. Vite + Onnxruntime + https://huggingface.co/Kim2091/UltraSharpV2
Now https://voluntold.fyi exists, and I never need to remember to manually move my single $100/year "ad free signup" off an event that has already happened to one that is coming next week.
Fun to build again, host on bare metal and all. Learned a lot!
The only criticism is that, in the UK at least, "slash" can mean urinate, and that's all I can think of when I read through the website. Probably a me issue though!
Spent a lot of time initially with asking it to generate me radically different approaches. Then I cherry pick what I like, iterate over and over again.
It's still something I ask of it regularly.
Thank you for the kind words. I've always struggled with UI as I have a backend engineer background, now more on the CTO sides of things. I've always lacked the creativity IMO, and this gave me a very good opportunity to practice and try to build something ""unique"".
All spamming with messages and news - so I made python scripts that use the apps API to get the messages (exams and changes etc) and puts it into our family telegram group chat if relevant
I just can't sit here every morning 20 minutes on my phone logging in checking everything just to find out the food plan has changed ...
Then, I got bored because they seem to be bad beyond certain complexities. But around Christmas last year, things improved a lot, and I’m getting confident building real ones. In the mid of these, I also got an offer to work with the [pre]sales/GTM team of a large company. So, I have been building working prototypes of bits and pieces of boring enterprise business around the world and have been racking in, if not billions, but very close to it in sales pipeline in about two months. (I did that 1B token in a week thing.) I’m sure the business and sales team will be able to convert 20% to 30% of that in the coming months. I also pitch and presented my work directly to customers and they are coming for more. So, it has been fun.
Lastly, I really wanted to scratch an itch I had for a while—build a Static Site Builder. But I ended up buildig it to be a documentation tool for TypeScript and JavaScript. It can be used as an auto-detect and let it build alongside front-end projects, make it a hybrid so you can add human-written documentation too or just build a Jekyll-esque static site.
O’Vellum is a 3-in-1 documentation tool. https://ovellum.oss.oinam.com
The work silencer/breaker is at https://void.oinam.com
The bubble popper that I re-created which my daughter loves https://brajeshwar.com/2025/bubble-wrap/
https://github.com/LukaWe/espCoinWatch - espCoinWatch, an ESP8266-based Bitcoin/Crypto Ticker with Weather Support.
https://github.com/LukaWe/LocalExifGeoMap LocalExifGeoMap - privacy-focused, browser-based tool to visualize and analyze GPS data from a batch of photos. generate interactive maps, heatmaps, elevation profiles, and trip statistics. etc.
- A Chrome extension to make design comments and ASCII wireframes right on the page you're working on and feed it to your AI coding tool as a prompt https://getdesignjam.com/
- An iRacing overlay to compare the telemetry of a reference lap against your live data https://flylapsim.com/
It's been a ton of fun to be able to dabble in software thanks to AI. Especially for small personal tools like these it's so powerful
I've also got a clang-repl wrapper for this codebase that is very easy to use and makes interactive programming much easier for me.
1. Comprehensive tool for auto expense logging and management, expense trend analysis, budget allocation, expense divider during group spend, report generator etc 2. DashCam app for simultaneously detecting threat and recording video. 3. Stock market portfolio management. A comprehensive tool which takes stock market investments as input, analysis the investments, provides a complete analysis, trends, suggestions etc 4. Fitness app. Records calories gained/burned,physical activities, all health parameters like Bp, Spo2, Blood sugar, Heartrate, Weight, Allergies and other synptoms. Analysis the health trends and provide suggestions regarding food, exercise and other health related anamolies. 5. AI learning app series ((13 apps to help learning AI from scratch) 6. Private chat app using Bluetooth communication 7.My own versions of Doc scanner, phone tracker apps 8. Health app which analyse tongue, eyes, and any symptoms to suggest the possible health issue and the remedy 9. JEE preparation CBT app 10. Electronic circuit builder and simulator app. (All basic circuits using AC/DC power, resistor, capacitor, transistor, diode, led, zener diode, switch etc can be created and run, it supports multimeter and CRO tools to measure different current/voltage and watch the waveforms)
Lot other things are in pipeline and will post once i complete them.
https://jaiph.org/
I then used it to build other tools -- a personal time tracker and a Wesnoth-like game (both not yet published). Basically, I maintain a Markdown file with a queue of tasks, and I run a Jaiph workflow in a loop that automatically picks up tasks one by one and develops them with no human intervention.
https://github.com/DavidSeptimus/alfred-jetbrains-launcher
Mostly, I use it to quickly open projects in cmux, but I use it for switching between git worktrees in IntelliJ too.
I looked for a decent remote keyboard app to use on the tablet, and found nothing I liked.
I ended up asking an AI to make something that served a webpage that connected back to itself via websockets. It provided a keyboard and touchpad on the webpage and forwarded events to uinput.
It works well enough on tablet and phone that I haven't got around to replacing the original keyboard.
https://football.sensecall.co.uk/
Other parents on the team love it. The live sharing is pretty handy when some aren’t able to watch the game.
Updating my machine used to be a dozen scattered commands: apt, fwupd, and a pile of dev tools I want kept current (node/npm, go, claude, opencode, plus binaries like kitty and lazydocker) that each update their own weird way. Now it's one Bubble Tea dashboard that checks each tool's installed version against upstream and only downloads what's actually behind, so I can see at a glance what's current and what's not. Adding a new tool is easy enough that I just toss new ones in as I start using them.
It also has a cleanup mode that hunts down all the pesky cache files that quietly eat your disk and reclaims the space in one pass, which has saved me from the "laptop full again" scramble more than once.
Currently, testing it privately and tweaking to make it awesome.
Another tool I made for myself is automated video editor (takes folder of raw photos images, and generates edited video that can be shared online). Used it among other things to edit all my GoPro raw footage laying around (hundreds of files).
So I made, kind of last.fm/waka-code for agents where I track (anonymous) usage per project
here is the example of my profile: https://clankerlog.ai/kodisha
Data is fully anonymous, all I collect is (agent, model, project name [can be mapped to something else to hide the true name]) and everything is opt-in by default (per project)
https://github.com/ityonemo/otzel
- Spector, a bare-bones CQRS library for elixir that composes extremely well with Otzel:
https://spector.hexdocs.pm/Spector.html
- nanodrop, an elixir library that interfaces with nanodrop spectrophotometers:
https://nanodrop.hexdocs.pm/Nanodrop.html
- opengenepool, a web-based dna editing tool (technically I started this one 11 years ago):
https://opengenepool.vidalalabs.com/ https://github.com/vidala-labs/opengenepool
- a proprietary camera library that communicates with industrial-grade digital cameras.
- a full on self-hosted lab management system (journal articles, obsidian-like notes, lab notebook, kanban, embedded dna editor, LIMS), obviously this builds on all of the above items
I think it was (a good idea), and AI made it easier for sure.
https://github.com/ncruces/wasm2go
- create your own coloring pages - https://www.coloringsai.com
- World Cup Prediction Pool - https://www.wk-pool.com
And many more
https://www.coloringsai.com/en/coloring-page/hitler-doing-a-...
I use Linux on my computer and an iPhone and wasn't happy having to use a cloud to sync my data between my devices. Syncthing/Synctrain + my apps allow me to keep everything on my private network, without ever touching a server or having to self-host something like Immich.
https://j23n.com/public/posts/2026/localios
OneNote to markdown/obsidian canvas converter. It did that using interop api to read the actual XML of the onenote files.
Work time tracker as 1px line on edge of monitor. Shows thin line at the edge of the display which fills up based on what i am doing.
Plaintext bookmark chrome extension that save links to local markdown file, Dynalist, Workflowy, Github Gist and import export between them. Was originally for Dynalist when AI couldn't do much 2-3 years ago. Recently added these other end points. [2]
A heart rate monitor with finger on camera. It's bit crappy though. Had to make it because many trackers, including google fit, couldn't detect 200bpm. https://github.com/SMUsamaShah/heart-rate
[1]: https://github.com/SMUsamaShah/LookupChatGPT/tree/claude/fix...
[2]: https://github.com/SMUsamaShah/plainmark
https://github.com/Opfour/coeus-ci - Named after Coeus (SEE-us), the 100 eyed Greek Titan of the inquisitive mind — whose name literally means "querying." CI stands for Competitive Intelligence. A business intelligence OSINT tool that builds company profiles from free public data sources. Give it a domain — get back a scored report covering stability, growth, tech maturity, financial health, security posture, and transparency.
https://github.com/Opfour/op4 - Op4 is a terminal-based encrypted messaging application written in Rust. It provides end-to-end encrypted private messaging with post-quantum cryptography, routed entirely through the Tor network so that neither the content of your messages nor your IP address is exposed to anyone — not even the person you are talking to.
Spartan (Private) - Open-source safety app for women. Community-based emergency response.
AATR - (Private) Catering unified platform: events dashboard, pack lists, and staff management
https://github.com/Opfour/thelinuxreport.com - Linux news and information aggregator.
God I love this stuff!
(edit: I also have about 6 more projects I am working on locally not yet uploaded to GitHub)
1 year later, with no js/ts skills at all, i got 10 custom plugins, several forks where i fixed bugs and some custom adapations, dozens of scripts and snippets and what not
Now obsidian works for me like I want, and still every day I use it Im still in awe
It is a bit hard to describe until I make a video but basically I seem to have changed the fundamental unit from a document to a chunk and it is just awesome
none of which would have been possible without AI
It's really removed all the pain points from my practice hours, especially when internet access is shaky.
- https://blunders.ai : Chess improvement app
- https://fretwork.ai : Freelancer management app (CRM/Billing/etc)
- https://validity.ai : Provide agents the ability to check the UI code it made (w/out needing to run through your full app)
- Save money on groceries + meal planning. This has probably saved hundreds if not $1k+ for our household at this point (some details here: https://x.com/ryanlanciaux/status/2063604299590939042)
- Orchestration / Starter Kit / Chat : Tool to help me manage multiple agent sessions at once. Some details on this one here https://x.com/ryanlanciaux/status/2063976049537417408
I realistically probably would have only saved myself less than an hour of crafting those expressions if the tool had already existed (with this level of detail, there are lots of many simpler ones already for it), but I would have spent a solid 40-50 hours of bouncing between manually crafting and writing detailed instructions to direct the agents to get this tool there.
[0] https://jsonpath101.com/
(Transcribes the audio, marks the timestamps, so you can delete a word in the transcript, and it’ll crop out that segment in the audio)
[0] Yes this is a plea, if anyone has the good stuff
I have a small RPI acting as my homelab pihole and dns so what better than to run the management UI on?! So I wrote a small bun management plane, nothing fancy, just a react app with user auth + openidconnect for those that like that stuff. From there, you have compute pool (empty at the moment because it requires a deployed agent). I added the ability to directly ssh into a machine, install the "agent" with privilege so it can manage docker, and the agent talks back to the management plane over websockets. A keep alive / health / status / resource packet every 15 seconds. Streams if you are looking at logs or accessing a container. I used Codex for most of this work but defined the protocol and everything upfront using protobuf (even though it's websockets). It helped with the "vision" and keeping the agent like Codex on the rails through completion.
Once you have a pool (agents installed on your N number of linux machines), you can deploy apps (which are my way of saying, a container with a namespace) or you can deploy agents (which is my agent, custom made for this) that are assigned to a project. I decided Org structures are a great way to delegate workloads so that's how they are modeled. Projects provide the git repo, the docker registry for images and storage of artifacts, as well as the history of all the prompts the agents have done in the project. Useful if you want to go back and search through |thinking| tags to figure out the reasoning behind a decision.
All of this was built in like maybe a month with Codex initially, until my agent was up to the task of coding w/ an endpoint configured (OpenAI API initially, now, NVidia DGX Sparks). What really works well is the delegation. The agent's have a webui that is exposed via the project urls so you can interact with the "scrum masters" of each project. They also share a stream if they are on the same project (but different subprojects).
I too wish there was more information on this but I didn't keep the lack of it from stopping me experimenting and finding what works. I came from the Mesos/DCOS era where you stop thinking about the metal and think in pools of resources. It's a distributed systems problem.
Few tools:
1) a combination of Python scraper and Claude skill to help family members find job by matching jobs to resume, to rank best fits
2) similar to above, but for stock data and financial news to identify movers and why they moved and see if anything is interesting.
3) a couple of attempts to import EPIC medical data from hospital into an offline app. Needs more work, data export from EPIC is crappy and a mix of images, pdfs, text, HTML and .jsonb files. Not useful at the moment
4) an application that downloads stock market data to run 15-20 strategies and back testing to identify stocks that match multiple and then run sentiment analysis on news feeds. Interesting, but semi useful. Needs lot more work.
Blocks the computer for x minutes. Agents keep on working. AI doesn't need a break, but I do. And honestly, at least for me AI has made my desktop as addicting my social media feed.
My own Alfred replacement is actually better for me, it's tailored to exactly how I want to use an app launcher / shortcut tool.
https://docking.cc
A Linux dockbar with tons of applets and support for x11 and Wayland. Works on Gnome and KDE. I always wanted to write one as I have been involved with several open source ones, but it is a lot of time to go from scratch. I use it everyday, and I am enjoying it so much!
A git worktree shell utility to quickly switch/manage git worktrees, and a neovim telescope picker which switches all the loaded buffers into the worktree version: https://github.com/jasonwoodland/wt
A terminal multiplexor plugin for neovim: https://github.com/jasonwoodland/terminal.nvim
A neovim session management shell utility (and telescope picker) to switch neovim sessions—no need for tmux or screen anymore: https://github.com/jasonwoodland/nvs
A NeXTSTEP-inspired menu for macos called NeXTMenus: https://github.com/jasonwoodland/NeXTMenus
A custom harness for pi agent which implements strict plan/review/implement iterative cycles and approval gating
A dynamic wallpaper utility for macOS changes the wallpaper when switching between light and dark mode — also useful for setting tiled bitmap backgrounds: https://github.com/jasonwoodland/macos-dynamic-wallpaper
A small macOS utility that dims the screen on a schedule, reminding me to rest my eyes: https://github.com/jasonwoodland/EyeSaver
https://plunio.app
Any parents with kids that cry in the night might benefit!
https://smartdomainfinder.com/
It uses an LLM to generate domain name alternatives that are relevant to your keywords, then checks whether any of those alternatives (in several TLDs) are available to be registered.
Warning: It's still a bit glitchy as I haven't fixed all the issues yet. It uses LLMs, but it's not a vibe-coded app itself. If it seems to be stuck while finding domain names, just refresh the page.
I also built https://github.com/lnenad/difiko as AI generates a lot of code that needs a nice way to review it.
1. family tree based on wikitree format. Transcribe records, verify/edit, then incorporate them into the tree with full citations and biographies. This one is the big one. It includes a tree browser and best genealogical practices.
2. Pool Math replacement. Log pool chemistry tests to markdown files and suggest the right amount of chemicals to balance.
3. Calorie counter. Log calories to a markdown file, look up foods and amounts in online databases, sync with garmin connect for exercise calories.
All of these are written with AI but also are interacting with pi and telegram, mostly using deepseek v4 flash.
It took more than a year. Why? Cloud infrastructure is too expensive.
https://github.com/michaelteter/docgen : create a single text file of your entire project, with a tree and some other useful bits. This is good for dropping into an LLM or research notebook instead of giving an LLM access to your actual project folder. It also can be put in your pre-commit script so you always have one single doc you can diff from one commit to the next.
md2pdf: markdown to PDF, relying on defaults and optional config files or cli args for formatting choices (such as page margins)
md2gslides: markdown, converted into slides, and using Google Slides API to generate the doc in my Google Drive. This saves me so much effort (I teach, so I make lesson plans/presentations all the time).
get-music: TUI app that lets me search Youtube and easily queue up to download one or more of the search results. Then I take the downloaded content, split out the audio, LLM process the video title, add metadata for music, and then provide an easy command interface for local searches and playback of downloaded content.
bookmarks: TUI for slurping all the URLs from my browser, LLM-tagging each url based on the tag list I provide with the prompt and url, and lots of features for managing priority, show/hide tags, etc. This was to help me stop worrying about having a hundred tabs open. Now I can just sweep them up into my own private, encrypted (sqlite) db.
ESL-Planner: Complete web app for building class plans for teaching English (based on params, such as student age range, skill level, specific teaching language (what we want to teach), etc. It's close to being ready to productize and release as SaaS, but I built it for myself initially.
Numerous other tools plus a guide doc listing all the tools and what they do. These resources are then made available to LLMs when I'm developing, saving me (and the LLM) the time of hand-crafting the same tooling over and over.
- Custom off-brand version of Pangolin
- Dashboard with beautiful UI for parsing traefik logs with database, filters, map and various integrations and statistics
- Samsung SmartThings Volume Control for Soundbar in Windows 11 native style
- Android App with good UI which serves as remote for switching display output modes for PC for movie night / gaming night with various toggles and for remote game streaming
- Many little one-prompt apps which run in background for QoL
- Reverse Engineering with IDA became a walk in a park
- A self-hosted comment system for my blog (https://github.com/karthikeyankc/discuss).
- A custom RSS reader with AI capabilities to keep tab of our competitors at work.
- A git-based CMS for my personal blog (which was also built with AI).
- https://www.yourfriendly.ai/. a desktop pet that lets me have an ai chatbox readily available
Vertical RTL by default, three paper modes (washi/sumi/ash) Offline reading, saved-list + per-position bookmarks
https://sutra-reader-3x2.pages.dev/
https://github.com/jessepcc/sutra-reader
https://theundercurrent.fm
* Highlight do-follow links.
* Spoof GPS - I live in a non-english speaking country. Sometimes Google sets my location to my gps, despite having an english vpn. This is an attempt to correct that.
* Local translate, rather than sending everything to Google.
* (non browser) An SSH selection screen, so I don't need to remember the IPs
> agentmux
> Configurable tmux agent launcher. Define AI agents (or any CLI) in TOML; sessions auto-launch the correct agent, tabs are colour-coded per agent, and prefix m cycles through the list.
https://github.com/lockyc/agentmux
I made it to deal with internet censorship in the UK, where different sites have different optimal exit jurisdictions, and most sites work fine so I can avoid the extra hops where possible.
It also works well for video streaming sites with geofencing, since the geofence itself is usually implemented in the frontend rather than the CDN. So only the frontend traffic needs to be proxied, while the bulk CDN traffic doesn't need proxying at all.
Socks5 is the ideal layer for hostname-based routing, since the proxy can see DNS names without needing to sniff TLS SNI (which would be incompatible with ESNI/ECH etc.)
iirc it was basically all done in a single prompt, and I've been using it ever since. The only issue I've encountered is with WebRTC not working properly with some services. (Presumably it breaks the NAT holepunching process or something, I never diagnosed it)
Another project that isn't quite finished is a "universal" web video downloader that works by shimming the MSE APIs and remuxing the streamed segments back into a regular video file. The idea is that if you can watch it, you can save it - including but not limited to youtube videos. I started this one pre-AI but AI was a huge help with the container format wrangling.
* Image viewer that can handle really big photos + run scripts via custom keybindings + CLIP search: https://daniel.lawrence.lu/blog/2025-10-22-sriv-simple-rust-...
It still needs a bunch of polish, but I use it a few times a week.
https://github.com/delecti/toof
* Codjiflo: A code review tool inspired by Microsoft's CodeFlow: https://codjiflo.net
* A virtual replica of a digital readout (DRO) for operating a CNC machine like a manual mill: https://el400.vza.net
* Reverse engineered CNC pendant integration with CNCjs also for operating a CNC machine like a manual mill: https://github.com/pedropaulovc/cncjs-pendant-whb04b-6
* A 'docker compose' to provision email, chat and documents for human-AI hybrid teams where you can take over AI's agent's credentials temporarily: https://github.com/vezzadev/roster
* The CNC stuff will come handy for a bigger project I have to create a 1:1 replica of Albert Michelson's harmonic analyzer: https://github.com/pedropaulovc/harmonic-analyzer
* Reverse engineered Hik-Connect P2P CCTV protocol for integration with OSS like Home Assistant and Frida: https://github.com/AlexxIT/go2rtc/issues/2289
* Some patches for different OSS projects like improvements to MCP tools, Playwright, Claude Code, etc.
While I was doing it I needed to render those diagrams as ASCII and I was surprised there's no Python library for Mermaid to ASCII. So I wrote one: https://github.com/fasouto/termaid (https://termaid.com/)
I've done a lot of little things in Emacs since. Just minor things to improve my workflows or build up Chief of Staff type information flows.
- Snubnosed mandarin app. Vibecoded anki and tinder-like character game for mandarin which allows new vocab to be added on the fly. Also accurate text to speech for tones.
- What did I learn? Tweet summarize that takes all favored tweets and assembles into weekly categories and allows deep research on certain topics.
A tool to periodically sync Device 42 data to Netbox (work).
Check sugar memory for the latest thing we were working on.
2. The second thing is when making changes across a large codebase agents are also dumb at figuring this out and also grep 300 things, using tons of tokens. Instead I say
Check RemembrallMCP to analyze the impact of the change.
1. https://github.com/roboticforce/sugar
2. https://github.com/roboticforce/remembrallmcp
* Auto-Birthday - if you have a contact in your android contacts that has both a mobile number and a birthday in their contact info, you can choose to send them an automatic "Happy Birthday" message on their birthday at a specific time. Can do it with hundreds of contacts. Doesn't use hardly any battery or resources.
* Wrecker - stupid simple "throw a ball at a tower of bricks and try to clear the board" game. High score tables. Made in unity. High battery when in use, No battery use when not playing. Will use internet for high score data.
* GeoNote - Create Geo-fences to generate a notification when you enter a location with your custom text in the notification. My wife is always telling me, "Next time we're here, remind me to only order one piece of toast" or something like that, so I make a note, it pops up the next time we're there and we're both happy. Notes are stored locally. No internet access required. Uses Geo-Fencing which is more battery-friendly than always-on GPS access.
All my apps are free, very privacy-focused and as battery-friendly as possible.
No information leaves your device (other than the high score data in Wrecker).
You have to side-load my apps though. I'm not putting them on the Google Play Store. They're so annoying to deal with! OMG
Nothing says “AI enthusiast” more than automating away social interaction.
It probably should message YOU rather than the person who's birthday it is, so you can send something personal.
Also coming up, automated wedding/funeral attendance by your personal humanoid robot designed to look and sound like you.
We already have that. It’s called a calendar app.
And a MCP-powered error tracking rails engine: https://dlt.github.io/blog/posts/mcp-powered-error-tracking-...
This was inspired by a karpathy tweet [0] and the prototype created using another tool of mine: The LLM Plugin Generator plugin (essentially a curated collection of plugins for simonws llm cli as a few-shot prompt)
The llm-model-gateway companion plugin lets you serve models from the LLM cli as a an openai API. This allows you to use saved consortiums in your various clients as if they where a regular model. Bringing massive parallel reasoning to any workflow.
It occured to me at some time that an collection of parallel LLMs was not really a consortium. A consortium is a group of organizations. A group of groups. To rectify this I added for actual consortiums, where each member of an llm-consortium can itself be a consortium of models. e.g.
llm consortium save cns-glm-n3 -m glm-5.1 -n 3 --arbiter mercury-2
llm consortium save cns-k2-n3 -m kimi-k2.6:3 --arbiter mercury-2
llm consortium save cns-meta-glm-k2 -m cns-k2-n3 -m cns-glm-n3 --arbiter cns-k2-n3
Yes, even the arbiter/judge can be comprised of a consortium of models, bringing parallel reasoning to the task of judging parallel reasoning chains.
Consortiums can also now contain groups of specialists. These custom user-defined expert characters address the prompt from a different perspective. And a Westworld style Attribute matrix can be randomized to inject some more entropy into the process.
[0]https://xcancel.com/karpathy/status/1870692546969735361
Some other llm plugins I vibe coded:
classifai generates labels with approximate confidence derived from logprobs
llm-alias-options saves inference parameters such as reasoning effort with a model alias. (good for setting the provider in openrouter or creating a consortium of high temperature models)
llm-prompt-json adds a --json flag to return the llm logs object (good for getting conversion_id, or reasoning output in scripts)
llm-jina adds support for all jina AI specialised models and tools like web fetching, embedding and reranking.
I think this is similar. Unfinished. https://github.com/mattjoyce/roundtable-consensus
Allows me to efficiently work on multiple tasks in multiple repositories concurrently.
Tuber[0] - this is my favorite, use it multiple times a week. It's just a little CLI wrapper around yt-dlp for my most common use cases - downloading the video, or the audio, or the subs. And then, if you've got the Claude CLI installed, it can also shoot the subs through Claude for a summary. I use it all the time, it's a great little thing!
Scrapio[1] - this is really specific but I was so pleased with how it turned out. You give it a list of "hacks" ("mods" for Super Mario World) and it goes out to SMWCentral, grabs each of the patch files, and patches a clean ROM. I think I only used it twice but it was just a nice way to chew through a list of hacks and get a few ROMs ready, made quick work of something that would have otherwise taken a bunch of annoying schlep work.
Lotus Eater[2] - calling this a tool is a bit undersell, but I'm still really pleased with it. It's a fan site for jamtronica greats Lotus that scrapes Nugs.net for setlist data and lets you do some mildly interesting analysis on things like song frequency and co-occurrence. Also has a per-user "shows I've attended" thing, Setlist Bingo. It's been fun to hack on.
Lastly, less a tool, more just a toy: last week Google released their Magenta model for doing live music generation. I thought it was really neat, and it's open source, so I opened it up with Claude, and after a few passes and some extremely annoying toolchain issues, I was able to add a spectrograph which does key / chord analysis to the "Collider" app, so you get a live readout of "what the band is playing" and you can pull out your guitar or whatever and join the jam with some info at your disposal. It's the kind of thing that would have taken way too much effort to be worthwhile in the past, but with AI, it's a really neat result of a fun night of weekend hacking. See the README I added in my fork for a screenshot:[3]
[0]: https://github.com/epiccoleman/tuber
[1]: https://github.com/epiccoleman/scrapio
[2]: lotuseater.epiccoleman.com
[3]: https://github.com/epiccoleman/magenta-realtime/tree/eric-mo...
https://bgremovefree.com/
- https://github.com/sethdeckard/loadout : TUI and CLI for managing a personal library of Claude Code and Codex skills across your machines.
- https://github.com/sethdeckard/atlas : TUI and CLI (also works as "cd launcher") that creates a smart, automatic map of every Git repository under your projects root.
I'm using these almost daily.
This one generates maps from OpenStreetMap data + some custom curated info in YAML: https://github.com/c0nsumer/trailmaps.app-map-generator
This one converts a basic chunk of OpenStreetMap data to an SVG so I can mark it up (by hand) in Adobe Illustrator to make specifically-styled print/PDF maps, such as what get installed at trailheads: https://github.com/c0nsumer/osm_to_ai
This one takes GPS recorded rides and builds custom/personal heatmaps serving up the map tiles so I can use them in map editing software: https://github.com/c0nsumer/local-heatmap-tile-server
And all of this has been put together to make the custom, local, specific-use-case maps that are at https://trailmaps.app (which, via local curation, are overall better mobile/online maps than many of the bigger auto-generated systems such as Trailforks, Gaia, RideWithGPS, etc, for visualizing local systems).
It's neat stuff where I understand all the inputs, outputs, and how most of it works, but AI tooling (Claude, mostly) has allowed me to bolt it together much faster than I would have writing it myself.
The idea came from using the Strava heatmap in JOSM to trace the proper location of mountain bike trails. I'm trying to use Strava less, and usually have ridden the trails enough myself before mapping them that I could use my own routes... So I figured why not have my own heatmap tile server?
It's also cool to just look at.
I could take it a lot further with time boxing what's displayed and whatnot, but generating the tiles is computationally expensive, so I just stuck with what I have for now. It meets the need.
On top of that, I made a Python parser that's meant to improve upon the awesome `orgparse`: https://github.com/Idorobots/org-parser
And now I'm building a CLI for Emacs Org-Mode, mostly focused on ad-hoc querying, agenda planning, etc: https://github.com/Idorobots/org-cli
I also built this site for educating potential customers and other privacy professionals about the increasing tempo of CCPA enforcement actions driving compliance: https://ccpa.world/enforcement
I could have probably coded this from scratch quicker considering that it took me two weeks to remove all of the hallucinated imaginary enforcement actions against real companies and also the citations to non-existent California law that the models kept injecting into my enforcement summaries.
Tubenote, a free YouTube video summarization extension. Mangata, a walking app that makes it easy to take notes and photos while walking. NotebookLM Clipper, a browser extension for importing content into NotebookLM. Knock, a notification tool that sends me a Telegram alert when Claude or Codex finishes a task.
and more products are also in the works.
- classless CSS library: https://leftium.github.io/nimble.css
- HN client: https://hn.leftium.com
- local realtime streaming transcription prototype: https://rift-transcription.vercel.app
---
These projects were started without AI, but heavily augmented with coding agents:
- https://weather-sense.leftium.com
- console.log replacement: https://github.com/Leftium/gg
- Thin layer over Google forms/sheets: https://veneer.leftium.com
As long as usage is not excessive, feel free to use the deployed version, as well.
https://donohoe.dev/subway/map/
- Meetup alert for meetups that match specific topics
- A daily journal that transforms entries into chibi-style cartoons
- A cashflow forecast our stupid accounting software can't do on its own
- DIY service monitoring for a ragged collection of docker containers, cron tasks, scripts and various others
The biggest unblock remains the tools/scripts/skills for documentation (started with Notion, network now sits on Obsidian for read/write).
1. A dashboard that tracks my personal metrics (github, strava, todo completion, flossing)
2. A eink display for that dashboard
3. A realtime node graph that shows a codebase (and/or its diffs) in a way that I can visualize what functions call which, and under what conditions
4. A agent that automatically fills out government forms and creates invoices for my friends brewery based on the delivery notes in their google calendar.
Runs entirely in the browser, no tracking, no analytics, no ads.
A tool that checks for new movie and tv releases, looks up ratings to see if they are worth adding to my plex server (see above about services I'm running), and then finds the magnet link and downloads them. But will only do so if my VPN is connected.
A tool that allows me to quickly build out paintball fields using my STLs of bunkers that I made, and export the full field layout as a single STL for quick painting, slicing, and printing.
most ambitious was browser extension to automate booking reservations since captcha and timing were critical.
swamp/evaporative cooler controller that monitors rate of change of relative humidity and sends RF signal to cooler to turn water on for x seconds to keep humidity lower than just leaving set to cool all the time. Does a great job but need to work on edge cases and ultimately replacing remote with my controller.
financial models for retirement planning
pen plotter gcode creator for old 3d printer to make labels.
food monitor to track what i ate with AI analysis based on symptoms recorded as well as interactive follow up questions
inventory app for electronic parts that now is inventory for all kinds of stuff like tools/plumbing etc that uses photo/ai to fill in data.
The wow moment came when it wrote syntax highlighting rules for 40 languages and file formats in ~10 minutes:
https://repo.autonoma.ca/repo/treetrek/tree/HEAD/render/rule...
imping - PingPlotter-like app. They didn't have a Linux version and I'm a paying customer, so I vibe coded this one: https://github.com/zenakuten/ImPing
utcolor - text colorizer for Unreal Tournament 2004 https://github.com/zenakuten/utcolor
utquery - Unreal Tournament 2004 Game Browser tool https://github.com/zenakuten/utquery
utstatsdb - This is an old project that did not work anymore with modern php+mysql. I had claude fix it. https://github.com/zenakuten/utstatsdb
https://guitar-tools.eejalab.xyz/
There are plenty of other tools already that do this but hopefully this one adds some quality of life niceties such as dark mode and a mobile responsive design that seems nicer to use than the others out there.
Hope the colours look ok. I'm a colour blind person (deuteranomaly) and I've optimised it for what looks good to me. Am open to adding a mode in line with what regular vision people might prefer.
No frameworks used, vanilla JS. No sign ups. Data/state persistence is purely local storage.
Absolutely no commercial interest here. Built this purely for the love of it. Will forever be free.
in case you forgot instruct your ai text generating tool to do so: replaces — with -, removes emojis and changes quotes to look like human-typed (even though they are not grammatically correct)
That way, you can who played without ruining the result. Then watch highlights in peace!
https://www.attie.app
It follows a lot of the conventions of Rails which is probably why it has turned out quite well
https://github.com/mbvlabs/andurel
I am going to develop the next version TermOnHost, which will let you connect to all kinds of hosts (Mac, Windows, Linux, or any Linux embedded system) And your hosts can connect to each other.
Since I live far away from family I also added an online game so I could play with them or show them what I was making more concretely.
I've cut some jigsaw puzzles that it made, but without access to an uv printer or a laser cutter that works reliably it's been challenging to actually make them
Jira Chrome Extension to add some notes and links to dashboard pages that I wanted for ease of use.
Small application which takes a CSV and turns it into a Registered Server List for SSMS in order to keep my list of servers updated for queries across all our databases.
Honestly one of the good use cases for AI. Small low complexity scripts and tools for assistance is a great use case. I'm amazed at the folks that are doing huge monolith rewrites with Agents and such, but I've never had good results for that. Small time saving scripts gets me a much more direct return on investment.
I write a Substack about the whole thing and have a pretty comprehensive list here: https://theautomatedoperator.substack.com/p/15-ways-im-using...
1) Use chatgpt pro from codex cli, opencode, claude etc as you can't get it via API. This has been the biggest boost in productivity for me as I don't have to copy and paste.
https://github.com/agentify-sh/desktop
2) A small gate to make sure any agent cannot run destructive rm -rf or git reset --hard commands, it has saved me many many times
https://github.com/agentify-sh/safeexec
3) For mac users, summarizes and speaks out loud after codex finishes a turn
https://github.com/agentify-sh/speak
Having worked with web facing teams there are always vastly different methods any individual uses to capture their feedback. If you or anyone you know on Mac that has 100s of screenshots on their desktop, this is aimed at bridging the gap.
Clipboards are optimized with context for LLM markdown ingestion and for use in work suites like Jira/Confluence.
Still fairly early, but I've been using the tool to help build the app itself which has been an enlightening experience.
https://third-space.astride.com.au/invite/c0378a6f-b1b9-4c26...
Bunch of security tools: Some are at https://diffsec.dev others:
https://github.com/diffsec/quokka
https://github.com/ihavespoons/hooksy
Revamped my blog to have a funky 3d background and animated cursor after years of minimalism: https://bdickason.com
A little screensaver inspired by After Dark: https://bdickason.com/static/experiments/flying-stuff/
A little toy using (mobile) screen tilt: https://qwertle.bdickason.com
A funky RTS designed for mobile: https://chasm-nine.vercel.app/
A start of a little 3D RPG: https://misty-woods.vercel.app/
Note: all experiences in varying states of completion ¯\_(ツ)_/¯
Highlights are that it completely free, no login required and works offline (once you "downloaded"/cached some files the first time around).
Currently working on a Gmail clone.
A web harness for another open source project (CHIRP) which lets you program channels into all kinds of handheld radios (HAM).
- I coded myself a portfolio manager to manage all the projects that I have
- secrets management tool to avoid accidental leaks of tokens by AIs
- tool for automatic creation of training/product presentation videos for web apps
- sales training app
https://archives.fifthrevision.com/color-generator/index.htm...
I also have a local zsh autocomplete macro that let's me type things like "git rename annotated branch" and ctrl-g and it will get me the actual command. There's also a ctrl-r mode that searches my history using natural language. This is connected to a locally run ollama so my keys don't leak.
- small club website: https://www.kolibrinkpg.com
- ticketing system with Stripe payments and QR scanning at the door
- Instagram/media ingestion for the club site
- genealogy tool with GEDCOM import
- scripts for downloading/archiving public-domain film material
- playlist/library tooling for DJ use
- music collaboration/sync tool for Ableton projects
- normal work stuff in a much larger existing codebase
I have become a lot more strict about process after being burned a few times. Mostly: make the change small, be clear about what it is supposed to do, check the assumptions before coding, use tests/logging/manual checks as evidence, and don’t merge anything I can’t review and explain myself.
https://github.com/dvelton/eyeball
It's called Commonplace: https://zby.github.io/commonplace/
https://mithraeum.studio - local first agent and editor in C, also a few models on HF (mainly jsut qwen wrapped atm but working on from scratch) https://fieldopt.dev - SaaS for dispatching jobs to the field (technicians, trades, delivery, etc.) https://github.com/zblauser/ytcli - youtube music from the terminal in zig (ps it’s free, no sub needed)
So ……. vibe coded.
https://github.com/pettijohn/corsair-ai-workstation-performa...
Pugneum, my static site generator based on pug/jade. Technically made many years before LLMs, but AI is fully maintaining it now so I think it counts. It's gotten to the point I believe it's superior to markdown.
The team behind Gel got acquired by Vercel and I already experienced falling in love with a dead database (RIP RethinkDB) so I decided to fork the concept to a TypeScript port with Svelte as the UI instead of React.
It'll live at disc.sh in a few months. Early dogfooding is promising.
EDIT: Also forgot that I removed React from GraphiQL in favor of Svelte too. https://news.ycombinator.com/item?id=48044888
- Converted invoicing to Typst (from LaTeX)
- Automation of blinds
- Automation of lights
- Python library to control lights
- ML tuning library
- ML feature interaction library
- Jupyter notebook slideshow interface
- Davinci Resolve Authomation
- Arduino eink bluetooth HR monitor
- Tons of small scripts
https://github.com/haydenk/homestead - another Go project, working on a better uptime dashboard that can also be the main homepage for accessing homelab resources.
I also used AI to find and create issues and milestones to for me to get the project to 1.0.0
https://azriel.im/disposition/
The text editor has an LSP built in to guide you to what can be entered, but see the examples.
I wanted a way to have visual documentation that can show/hide detail, and be pure svg so that it can be pasted into sites without becoming pixelated.
Also supports animations to show interactions between hosts, which always gets messy when drawing a static diagram.
Can be used to resize images, but the main purpose was pixelation for a game I was making.
Calibre web UI 2.0 to replace Calibre's mediocre web interface. Used for browsing ebook catalog, searching and cataloging with a simple list feature.
Project Gutenberg local mirror UI to browse my local copy of Project Gutenberg books
A couple of MCP servers for self-hosted services to give access to OpenClaw. Currently working on a daily digest that Claw will generate that includes feeds from these: what news stories were popular in my feed reader, did my baseball team win, etc.
https://github.com/btucker/graftty
token speed calculator - for estimating tg/s of ai based on ram speed and model size/params this helps in comparing different hw, estimating likely speeds i will get on hardware
prompt assembler - to create prompt and context once and reuse it in different ai's, picking and choosing context in a prompt, creating agent.md etc.
dashboard builder - for viewing gsc, ga, stripe data in one place
Being proud of the result.
THAT is a real game changer LLMs allowed me, both in my professional and my casual life.
For example this:
https://github.com/yodalf/coincan.git
or this:
https://github.com/yodalf/kiosk.git
I can once again write Clojure apps for my phone, which is fun to do by hand, unlike more conventional tools for writing Android apps.
Among many others
https://github.com/hn-ai-podcasts/browser-extensions
Code: https://github.com/VadimKey/xorpingtonian
Catalog (in Russian): https://vadimkey.github.io/xorpingtonian/
During vibe coding I found that emojis are not that simple as I thought about them.
https://github.com/gitsocial-org/gitsocial
1. A Mecrisp-Stellaris Forth LSP for the Helix editor (Python) 2. A CMSIS-SVD Sqlite3 search and paste pop-up window for Neovim when editing STM32xx embedded Forth programs. (Lua)
https://mecrisp-stellaris-folkdoc.sourceforge.io/
https://www.motormait.com/
Bc my version uses simple copy/paste rather than deeper OS integration, it works more reliably in the Claude Code terminal (has to do with active windows or cursors or something). And bc it uses local Whisper, I also find it transcribes technical words such as “git” more accurately.
Nothing technically challenging but practical for me.
Will improve the read me
After that I started writing opub.dev because even minimal success in recent oss showed me just how much has changed, and I’m worried about how expensive everything will get for maintainers.
So, now I’m trying to GIVE people compute so they can start building a helpful filter layer above their projects.
Automatically rename screenshots: https://github.com/amichae2/screenshot-renamer
https://bedtimebookhelper.com/
After coming back from paternity leave, I found that my team had really leaned in to AI driven development. This project was half catching up and half attempting to solve the burn out from the repeated books my wife and I were experiencing.
I've also built a release notes app for my QA teams, a DORA metrics app, a thing to map UX journeys with Playwright, and a ton of games and stuff. AI got me back into enjoying building things again.
FOSS https://github.com/klinquist/Notesync
Screenshot here: https://x.com/LyleMakes/status/2063784301594853657/photo/1
https://github.com/jeffnv/lockin
No code or docs was hand written for this one.
But no honestly, unfortunately most tools I did for myself are not for hobbies but something that I needed for work... like this one (https://github.com/crmkit/crmkit) most recently.
- A chat-based web app for ad-hoc telemetry data visualization. - A firefox-extension meeting transcriber - A personal chief of staff - A web-based personal finance budgeting tool (no AI at runtime, obviously) - An iOS and Android app for solo work with a dead man switch if timely check-ins don't happen - A custom dotfile/machine config manager that works the way *I* want - A bookmarking tool/web clipper that puts together daily content-only collections from what I've been clipping and send them to my Kindle. This one I actually intended to make a SaaS, but meh. Being able to put together quite big projects by myself for myself in a reasonable amount of time is such a joy.
A remote image viewer to see screenshots in VMs.
A simple agent harness to drive spec to verification.
A YouTube video summarizer.
https://github.com/ozten - some public repos, but the majority are private repos
Somewhat related - I wish there was some local thing I could give my 100 holiday videos and it made something fun with the highlights to a specified duration.
1) Automatically editing out pauses 2) Making those TikTok-like captions
It has "levels" of compilation, with EASY mode being about as easy as Ruby, and the compiler can present you with options to get that as strict & performant as Rust/Tokio.
I'm going to need at least a month to finish all the documentation, though.
https://mediaden.ca - iOS app for storing encrypted photos/videos on storage I (the user) exclusively owns, with zero servers, zero telemetry, and a host of other privacy related features.
https://github.com/klinquist/machinemon
The final idea is to own all my data, but I’m still on it.
Pretty happy so far
This started off as a fancy cron with webhook and became a comprehensive runtime. I have been using it for months on several systems.
I use it, and have given my students access to it too - they use it to help their revision.
The Dead Classroom Theory.
Static site generator for my blog, or at least bits of it.
Already have $180 ARR prebooked (the money that I used to pay for youtube music), looking forward for more.
if anyone has links for open-source self-hosted spotify/yt music replacement, I would gladly appreciate links
Github: https://github.com/jantznick/youtube-spotify
It essentially uses youtube as the music source, I think I heard somewhere that playing through embedded videos skips ads but I'm not really sure, in all my time testing it I never noticed ads, but I'm also on premium so that may have been why.
by all means critique, I don't know that I have a ton of time left for it and I'm sure there's bugs here and there. I was having issues getting it to autoplay on desktop when the window itself wasn't the active tab. I never really tried it on mobile.
I was trying to get some DB of artist/song info but doing that was proving to be complicated.
I collect song metadata from various places (genre, instruments, track credits, rating). I also scrape charts by year, genre etc.
Then I run an ETL job on the json data I have downloaded, pre-building queries for extremely fast lookup tables. This gets saved to Duckdb, which is used by my go web ui/api.
It's very early days, and I only spend one or two hours a week on it, but right now it's amazingly useful. It had roughly 80k song metadata. To preview the suggested songs I ended up building a very cut-down YouTube music player, except that the playing song has all the metadata right there, and everything is a link that can take you to the artist, composer, instrument, genre, album etc. It's a great way to "wander through your collection".
Unfortunately this is only useful to me, because I targeted the music I listen to.
Next step is to download lyrics and extract song meaning, keywords etc. Then use MusiCNN, (or CLAP,OpenL3, HTSAT) to extract embeddings. Finally train my own model for nearest-neighbor retrieval based on a mix of metadata, giving the user the ability to tune it on the fly.
For now I'm letting that backend access my files directly. The front end can also play YouTube music (free, using the yt-dlp method).
None of this is public yet so right now I don't need anyone's approval.
https://packagecontrol.io/packages/Claudette
It has less features (no OG media or title/story analysis via Bedrock) but it focuses more on the features I like/need from an HN client
https://buckets.joelryan.com
https://github.com/mohsen1/yek
It wasn't perfect, but it helped me feel confident in the final result.
ive made some wallets, a incredible pair of ear muffs, and a bunch of key tags.
i keep being asked if im selling anything, and when i get the next piece together of building/buying an embroidery machine that can work on leather, i might
i still dont want to buy a proper fabric cad system, so im trying to figure out a minimal version for making glasses cases for everyone's christmas gifts. its handy being able to draw inputs for claude, but its also nails-on-a-chalkboard
https://sisuonspeaks.com/
I wasn’t planning on posting it yet, so I’ll have to get you a thorough follow up, but in a nutshell it’s: a continuously running 448-concept space (philosophy, cognition, art, nature, math) that occasionally “crystallizes” a group of 4 concepts via Hebbian learning and stochastic noise. Those concepts get sent to the LLM with minimal guidance beyond some safety guardrails and encouragement to be creative. It has access to a sandbox to produce essays, stories, music, “art”, and small python scripts. Self-updating memory system. Notes, essays, and artifacts can also be discussed with me through some outbox channels.
On top of that it’s got a separate academic philosopher + psychiatrist llm that critiques its work and has a regular cadence of “sessions” with it, as well as a research assistant bot who I talk to (but doesn’t interact with Sisuon) who has full project context and memory access. The sisuonspeaks site is a VERY abridged collection of Sisuon’s essays, along with analysis, commentary, forum posts, and a podcast…all created by, you guessed it, more LLMs.
Eg: push a button, it shows that it's working for a while, then strongly flashes when it's done (success/failure). When you have it right under the monitor, it's like a macro pad for long-running things.
This reminds me of some of the very early peripherals you'd see on the Alto and other computers. I was surprised something like this didn't seem to exist, but maybe I'm just terrible at searching.
https://github.com/ericfortis/tabular-eye
But I dont use "AI" to make them
I use a code generator
I like to use the smallest possible "toolchain", using the least possible resources, to build software tools
Ideally I want the tools to compile quickly on underpowered hardware
https://tools.dsebastien.net/
https://charleswiltgen.github.io/TagLib-Wasm/ – Also built for myself initially, I FOSS'd it because there was nothing like Mutagen for TypeScript/JavaScript runtimes. (I don't dislike Python, but think it's a bit of a mess.) This was my first serious project to leverage LLMs for coding.
https://pwascore.com/ – Built because I wanted to quantify how bad Safari was at PWAs. Learned that, objectively, Safari is as bad as PWAs as Firefox (which is to say, not terrible, and not to blame for why PWAs continue to be mostly-irrelevant).
basically trying to see what a vertically integrated agent looks like, where the agent has deep access inside a framework and it operates from within a framework, so like, instead of reading files, opening processes etc - it gets a bunch of framework specific runtime tools(logs are the easiest example)
When you install the CLI, it (with your permission) asks to update your base agent prompt files (e.g. `~/.codex/AGENTS.md`, or `~/.Claude/CLAUDE.md`) with info about how to use the tool.
This means all your agent chats know about SDocs, and it’s nearly always your agent which invokes the tool: “Hey Claude, sdoc me a list of all my open MRs”, etc.
I did a ShowHN about it here: https://news.ycombinator.com/item?id=47777633
I also built a new web framework we use internally which is amazing. We might open source it soon. It has a postman clone that has a bunch of features I wanted. It really is the case that we can just build tools any time we want.
https://github.com/TrevorS/rmux
Also curious if you know anything about scraping Twitter.
For reddit there used to be the json endpoints that you could just fetch, and you can batch your subreddits, so its nice end easy. They have just killed those...old.reddit still works, but i fear like the days are numbered there as well.
Reminder.dev - Quran app and API that includes RAG search to provide a more authenticated source of summarisation. The first thing I dabbled in with AI.
Micro.mu - Rebuilt my entire product idea from 10 years ago as a super app for daily digital habits. Something I use everyday for myself.
Aslam.org - An islamic knowledge base that uses sources of data as a way to ground AI chat and make notes. Very useful tool that I'm using on a daily basis.
Go-micro.dev - Totally revamped the open source project, docs, etc with Claude.
Its open source (https://github.com/micro/mu).
It really helps us to not forget to spend significant time with each other when life is busy.
A rant follows.
I've generated probably as many lines of code by this point as I've written myself over the past 5 years or so.
I found AI generated code mostly very frustrating, kind of low quality in its own way, and too complex. I have pages and pages of instructions to guide the agent(s) to do a better job at this, and it has gotten better, but the fundamental limit of this technology is tangible.
Like, okay, CPUs still get faster every year, and every now and then someone makes a breakthrough and we get a bump in speed from something. But when you write high performance code, you very quickly run into hardware constraints, like how fast information can move and how far away components are from the CPU cores themselves. People keep saying performance isn't THAT important, and that modern hardware is so fast and amazing, and that they struggle to even find a way to use all of their CPU cores and RAM with their little app or program or game. Yet here I am, writing code that will noticably speed up if I run it on a CPU with a little bit more L1 cache.
This is similar to how it feels to program with AI when you're reasonably competent (to put it mildly; I avoid the 10x developer label because comparison to others is very silly). Everyone keeps saying it's getting so much better, and it's so good, and worrying about code quality and architecture is dumb because we can move so fast it doesn't matter. Yet here I am, writing code by hand because I tried doing it with AI a couple times and it just doesn't hit the mark.
I'm not doing anything special, I just have high standards and a good amount of experience when it comes to software quality, performance, and maintainability, which is why I keep getting hired. I'm convinced that people who think their AI generated software is good are the same people who write short variable names and think it makes their software faster (hyperbole, but you get what I mean).
I can feel when I hit the limits of the hardware, and I can feel when I hit the limits of LLMs, and I know for both of them that a 2x increase in performance will not change what is and isn't physically possible.
Art search for magic cards
https://github.com/yen223/hotpot
Side note: before coding agents I would not get passed the branding and login page.
I needed to see health of many Function Apps and Container Apps in a single page
https://github.com/RobbertH/azpect
- a slop detector / browser extension that filters slop replies from twitter/hackernews/reddit: https://slopsieve.com/
- tweethoarder ( https://github.com/tfriedel/tweethoarder ), saves my liked tweets and makes them searchable
- mattermost_archive - syncs all my mattermost channels and makes them searchable via an MCP in claude
- https://github.com/tfriedel/asana-exporter - same thing for asana
- https://github.com/tfriedel/dynalist-archive - same thing for dynalist
And the inverse as well, of course.
Runs on a raspberry pi that I was otherwise using to take backups periodically… has been working pretty good honestly.
I also built a program that fills/submits my time reports, and does the same for all of my subordinates - then signs them off… Saves everyone like 5 minutes if we remember to do it, or 12 minutes and frustration if we’re reminded by HR about it (which happens because who the hell cares about time reports?)
https://convert.neocities.org
https://github.com/mretallack/OpenEPaperCliTool
----- 3d printer pipeline, so its can print stuff directly without having to use the computer to set it up.
https://github.com/mretallack/3dprinter
----- Experiment with creating a Abdroid Auto app for phones that cannot run real AA. (WIP)
https://github.com/mretallack/AndroidAuto
----- A android 3d clay modeler to create models for 3d printer, with stl export.
https://github.com/mretallack/ClayModeller
----- Uk Fuel finder python lib and Home Assistant intergration for showing fuel stations from UK gov api.
https://github.com/mretallack/ukfuelfinder https://github.com/mretallack/ukfuelfinder-ha
---- Reverse engineer cheep drone video feed, from drone found in charity shop
https://github.com/mretallack/DroneCamera
---- App to send voice to camera using mqtt.
https://github.com/mretallack/CameraSpeaker
---- Added ONVIF to an oss rtsp android app.
https://github.com/mretallack/cams
---- Added Home Assistant to Dicio Assistant.
https://github.com/mretallack/dicio-android
---- Added telegram bot interface to kiro, with group support.
https://github.com/mretallack/kiro-remote
Deepseek v4 pro does a pretty good job of actually adhering to the word restrictions.
Most language learning content is "slop" anyway -- so might as well generate slop that's at least a little interesting.
Presentable: a photo library sorter with Ai powered organization assistance, a compare canvas with various viewing modes and customizable folder sorting shortcut templates (wip).
AlwaysWhisper: a tool that let's me attach STT to my entire OS, with custom wrappers for different programs, adding theoretical voice control to any software (wip).
ScreenLoader: an Electron based tool that can load any web source as a kiosk app, full of useful features like keep-alive, covering multiple screens and tracking output logs.
Inputboard: a unified all-inputs hardware board, that transfers input data to any prototype I want to work on using an optocoupler, so I won't have to fiddle around with setting up clean and reliable inputs from cheap Chinese components every time I just want to test something.
Squire: an agentic board game helper, that can ingest a manual and will hopefully help decrease the time spent on endless discussions about seemingly conflicting rules. It should also be able to help me play a game when I don't fully understand the rules myself yet (wip).
NodeRunner: an agent that plays the WikiGame, focusing on speed, efficiency and token usage (the result of a fun competition with a colleague).
Sonic Bloom: more of an experiment than a tool. It's a wireless piece of custom hardware, that listens to conversations, sends data to an LLM through fast STT and returns a color choice that matches the topic being discussed to the hardware, which then controls an LED ring. It also has a small display that explains the logic behind the color choice.
Image-to-story: a VLM tool that kickstarts a written story using an image and has some rudimentary tools to expand on it based on user instructions (wip).
At-work-or-not: and Android app/website where colleagues can check if I'm working from home, if I'll be at the office or if I'm not working at all. Also doubles as a private record for tracking transport expenses.
SharedMaps: a Maps based website where groups of people can share custom categories of geo locations and drop comments on them.
VMG: an image format that includes audio with images and offers TTS input to easily add narration.
Who wants Coffee: a small Android app to help me remember who wants to drink what when I go for a round at the office.
And a pile of Python scripts for smaller useful tasks.
Nothing major, and only works with my infra, but it saves me a few minutes a few times a day to just be able to check the tab, and if there's an alert load up the full stats page.
- I didn't like that I can't use my newsreader on my laptop and my phone as easily so I built https://github.com/mjc/nntp-proxy. that turned out to be really hard to benchmark once it got fast enough so I am working on an nntp benchmark tool https://github.com/mjc/nntpbench. both can do request queuing because the nntp RFC says servers have to accept as many requests as they can, and then process them in order. so if your client doesn't do that, you can use more connections to the proxy and it will queue for you. it also routes stateless commands to whatever server is least-loaded, and will switch to stateful mode if your client needs it.
- I didn't like how expensive AWS Transfer Family is, so I built this https://github.com/elixir-ssh/sftpd and then rewrote it in rust (alpha) https://github.com/mjc/sftp-s3-rs. this shook out a bunch of bugs in russh, which was fun. - didn't like that there's no par2 implementation in rust so I built this https://github.com/mjc/par2rs (I'm too lazy to move to tape backup so it works pretty ok for dvd/bluray parity), unfinished but good enough for my use. - same deal for 7zip in rust. https://github.com/mjc/r7z - a medication tracker thing that uses claude/codex/copilot to scan the bottles and parse them as well as identify pills etc. works better than you'd think but I'm not planning on releasing it for a while.
fixed or fixing bugs in: - exqlite (it should not crash anymore and should return busy a lot less often.) - russh - swift-nio-ssh (this might be why codex's remote can't connect to your ssh box) https://github.com/apple/swift-nio-ssh/pull/236 - NanoKVM (working on making the streaming for this a lot more fluid)
2. A shopping list app that allows me and my partner to coordinate on our shopping
3. A recipes app that includes AI scanning
4. A standalone home battery dashboard/app
5. A fuel prices app that is tailored for the closest fuel stations and is ad-free
6. A tool to draw classroom supervision maps for my partner (thrown away already, I didn't want adware/bloatware so I built it, she used it, then I threw it away)
7. A quiz website, cos the one I used to play on was overrun by ads.
8. A time tracker that I'll throw away at the end of the tax year
And more, and that's just what I did for making my life easier, there are other more "enterprisey" things I am working on. They're web apps that I add to my iphone desktop or run on otherwise junky old tablets or on TVs.
The point is that they do exactly what I want them to do instead of relying on downloaded apps that get me 80-90% of the way there, even if they'd be classified as "AI slop". I know enough about security and caching that they aren't full of holes and don't kill upstream, but I don't really care about the code, and it's literally easier for me to build something new than to go to Google or an app store to find software that's full of ads.
- gcode sender and generator https://mycnc.app
- CNC simulator https://sim.mycnc.app
- Cabinet design with door/drawer designer https://cabinet.mycnc.app
- a stateless dashboard for work that collects from 6 other APIs
- a refactor of a huge function with 8-deep indentation into readable small functions
- a road trip game for my kids where you take photos of things from the car
- github clone + extras
- a stack (FILO) based task manager / TODO list
- a CAD kernel with Blender frontend (WIP)
- a minecraft mod that makes real terminal emulators in block form
- ^ that but in Godot + a terrible "game" world (WIP)
- a somewhat failed app organizing claude workspaces
- a somewhat failed attempt at a VM framework for MacOS
- a somewhat abandoned gmail clone
- a farmland pricing model + maps etc.
- partially reverse engineered VCDS device
- a likely novel fractal system I need to work on some math to publish
- NTSC transmitter/receiver in gnuradio for the artful corruption of video
- backend for iOS appstore handling of account/subscription things
- an RSS / Podcast reader
Just a more helpful discord chat generally. It also gaslights you too!
Here is the tool: https://git.sr.ht/~asibahi/hoopoe
Currently working on a markdown search and wiki backed by Typesense, also has good web search, fetch, crawl. This will power my personal knowledge base system as an important step towards more leverage and better outcomes.
https://github.com/verdverm/gmd
Here is a website I made with places.js for DC area board game events. https://dmvboardgames.com/
Built a meeting-intelligence pipeline that turns raw, error-prone transcripts into a structured, queryable knowledge base. Meetings get auto-transcribed by Krisp, whose speech recognition mangles the things I most need correct, like colleague names, customer names, internal product and architecture terms. I hand each transcript to Claude alongside a hand-built context document, and it works a fixed routine: read the context file, read the transcript, then reconcile every uncertain name or term against a master error table before drafting anything. Only the genuinely unresolvable handful surface as questions; everything else is corrected silently. Once I confirm those, it emits a cleanly formatted markdown summary in a manner I describe as a template: overview, topical notes, decisions, action items — and pushes the work items into Todoist so commitments don't get lost.
What makes it more than transcription cleanup is the back end and the feedback loop. Each summary hits Obsidian with YAML frontmatter and live Dataview queries, so open action items and meeting metadata behave like a database rather than static notes. In Cowork the whole accumulated Obsidian folder becomes fully queryable rather than merely searchable — instead of grep-ing for a keyword, I can ask questions that reason across months of meetings ("what were Todd's table-stakes asks, and has anything shipped against them"), with the model able to look across separate conversations. The other half is self-improvement: every clarification I resolve gets written back into the context document: its people directory, terminology glossary, and especially the ASR error table, so a garble I corrected once is corrected automatically from then on. Over time that one document has become a domain-tuned lens, and each meeting both draws on it and sharpens it, which is why the summaries keep getting tighter and need less of my intervention.
Beyond that: I use a Netatmo weather station which has a RESTful API (or sends to a cloud server that has one) - I pull that information (which I can see on the web and their apps) into my own VictoriaMetrics / Grafana set up on Kubernetes, via a Go app Claude built.
This app above was when I had my little aha moment: Netatmo's OAuth is slightly broken (issues with the different tokens and refresh). But I'd written a dog-ugly app which managed to work a while ago. Claude kept trying and strugglign to understand why its OAuth code wasn't working, and was asking me "are the credentials right?" etc. "Yup, I'm able to get data from my old app", then it said "If you have the source code to that app, I can figure out what's up", it looked, identified the issue, tried to work around it and then we "agreed" - "Hey, this should work this way, but it doesn't, and whether my old OAuth code should work or not, it does, so drop that in, and keep going". "Great, let's do that."
Just a super minimalist thing where each day is one .txt file with the newest one at the top and a lazy loading scrolling list with every note going back 5 years
Supports CTRL + F searching, backups, and a bunch of other QoL features/macros
Its kind of a revolution that with agentic coding everyone can have their own hyperspecific customized apps
I find a phone call is more likely to get through to me than a reminder or alarm, which I can ignore or forget; an ordinary reminder is not as interactive.
Claude built it all and although there's a script for it, I just set the reminders in an interactive Claude code session in the directory. (Like I'll open a claude code session there and say "using the script in this directory, call me tomorrow at 7 a.m. with the message 'dr's appointment'."
It works well for me.
Would have made them without UI with a bit more elbow grease invested in web-searching for some examples, maybe even a StackOverflow question.
Generally, I'm not a fan of LLMs and their social effects.
It's a user daemon that runs on my machine and exposes a unix socket, and then a bunch of hooks in claude, zsh, vim, etc, that report directory and commands I've run and all that, pipes it to claude Haiku for summary, and then stores context in sqlite. It also exposes that data as MCP so I can use claude to say "hey what was I doing yesterday," or any arbitrary time range.
I find that in the age of using AI agents, "Wtf was I working on yesterday" is an even harder thing to remember for me, so this helps me kind of track everything with a database that a) has AI summaries already and b) can be accessed by AI as well as a CLI.
https://github.com/clashleyca/magpie
Browser emulators / games
https://nesvibes.tsilva.eu/ — Browser JavaScript NES emulator in ~2.5k LOC
https://scummweb.tsilva.eu/ — Run ScummVM games directly in the browser
https://github.com/tsilva/REFramework-chill — Anti-nausea improvements for Praydog’s Resident Evil VR mod, specifically for Resident Evil 7: movement vignette, snap turning, etc. WIP.
AI / ML browser experiments
https://llame.tsilva.eu/ — Run small LLMs in the browser with WebGPU
https://aipit.tsilva.eu/ — Debate simulator: pit simulated personas against each other
https://aigrounds.tsilva.eu/ — Playgrounds for experimenting with math, AI, and ML topics. WIP; lots of slop and most are unreviewed.
https://modelviz.tsilva.eu/ — ONNX model graph visualizer
https://modelarchviz.tsilva.eu/ — View model architecture diagrams, codebase, and paper side by side. Includes a chat bot that can see selected content and select content itself. WIP.
https://embeddingviz.tsilva.eu/ — Visualize model embeddings and/or layer activations for different content with PCA, UMAP, or t-SNE. WIP and poorly tested.
https://github.com/tsilva/dlab — Workbench for deep learning experiments. Open the repo in Codex, pick a training target, e.g. maximize validation accuracy on CIFAR-10, and ask it to start a research track. Then collaborate on training runs, evaluations, sweeps, and iteration toward the goal.
Datasets / visualization
https://minariviz.tsilva.eu/ — Minari dataset visualizer: https://minari.farama.org/
https://github.com/tsilva/gymrec — Record and replay gameplay from Gymnasium environments as Hugging Face datasets. Supports stable-retro environments such as NES, SNES, Genesis, etc.
https://github.com/tsilva/youtube2datasets — Convert YouTube videos to Hugging Face datasets.
Agent / coding workflow tools
https://github.com/tsilva/runbook — Run Jupyter notebooks on Modal through a CLI, with streamed outputs.
https://github.com/tsilva/agentpong — Run coding agents in VSCode/Cursor terminals, one per Aerospace window. Agents trigger desktop notifications when done; clicking a notification sends you to the correct desktop. I haven’t used this in a while, so I’m not sure it still works.
https://github.com/tsilva/agentbox — Run agent CLIs inside Docker sandboxes. Project-level config defines sandbox access. I stopped using this once agents got better built-in approval mechanisms.
https://github.com/tsilva/agentbridge — Use an agent CLI subscription as an OpenAI-compatible API server.
https://github.com/tsilva/claudesk — Agent coordinator for Claude, built before decent agent orchestrator UIs existed. Stale and probably broken now.
Personal workflow / GTD
https://github.com/tsilva/gmail2obsidian — Flush labeled Gmail messages to Obsidian.
https://github.com/tsilva/thunkd — Quick idea-capturing mobile app built with React Native. Sends notes to Gmail, which is my main GTD inbox.
https://github.com/tsilva/capture — Quickly capture thoughts to Gmail using an Alfred shortcut, for my GTD workflow.
File / document utilities
https://dedrive.tsilva.eu/ — Client-side Google Drive duplicate finder. Lets you choose which files to keep. Poorly tested; use with caution.
https://github.com/tsilva/pdfpress — Misc PDF tools: split, merge, compress, unlock.
I'm more than a mere typist; there's a skill to getting a usable, useful app out of an AI.
There is no skill to using LLMs, you can learn to be effective with them just by using them for a few days.
Even babies can learn the technique!