It's very cool, but it's also a shame that measuring in characters leads to the metagame of using 97 multi-byte Unicode characters that decode to 194 ASCII characters when reinterpreted. Almost everything is in the format:
Can't they just have the same agreement that Ford Prefect had with Mr Prosser? "So, assuming I'm going to use this lossless compression technique to fit 194 characters into a 140 character tweet... how about you just show the 194 characters and say that I fitted into 140 characters?"
I suppose you could make the same argument about size-based competitions at demo parties; the limit is strictly 4096 bytes but everyone + their dog is using Crinkler to compress a 12-20KB executable down to that size. In fact, part of the effort is in aligning data and tweaking constants with Crinkler's algorithm in mind, to make the raw data more amenable to compression! But at least then, it's not a constant compression (turning any 194 ASCII characters into 97 Unicode characters which with embedded decoder makes exactly 140 characters)
The Sampler contains all of the images in the html file and still comes out to be only 13k. Probably quite a bit less if you removed the ray tracer and scenes.
i went super deep on 140-byte code golfing[1] back when twitter was taking off, and it changed the way i think about code. for a brief while we had a small community collaborating on byte-squeezing techniques[2], and i was constantly amazed at the creativity that this constraint brought about, from mandelbrot renderings to sudoku solvers. possibly the best part was more than a decade later when i found my golfed UUID implementation[3] deep in my employer's codebase.
I love these but I wish they'd disallow the eval thing. I also which they'd added more shortcuts. Like IIRC they have `s` for `Math.sign` but so if they're going to shorten things they could have gone much further. `globalCompositeOperation` etc...
You have to keep in mind that for anything like this, the platform has to exist before the ways to game the rules are thought up. To change things after the platform is up and running makes it a moving target and for something like this a fixed target is intrinsically part of the appeal.
Dwitter used the definition of what Twitter at the time used. The solution of https://beta.dwitter.net/ mitigates the lack of accessibility of the encoding, while keeping the fixed target at the same time introducing a new practical target.
The Math.sin css color encoder were concessions to practicality, there has been much discussion as to whether a Dwitter 2 should simply include additional characters required to the same preconfig and let people do it themselves, and let them use the characters for something else if they can think of something.
I'd like something like this just to see how creative people get in doing that. How small can you make a bit of JavaScript that copies every static method of Math to window?
and creativity is the goal here. No matter how people stretch the system, it can't be denied they have been doing so creatively.
You have to fix the scoring. Blacklisting eval is whack-a-mole forever. If you switch the scoring to use bytes of UTF-8 encoded text, that pretty much aligns everything. There might still be some use for packing data in string literals, but wholesale source code packing would probably mostly stop.
Yea, because no one has ever changed any rules. Every sport has changed rules, lots of games have chagned rules. It's kind of irrelevant. Removing eval and adding new shortcuts would just start a new era of creativity.
OT: we should really extend bbcode with a safe scripting language. The hilarious stuff that could happen in threaded discussions. You could have a sub set of safe instructions that get executed automatically, a larger set of instructions that require pressing a button under the comment. (And a button to view the code) Then a 3rd set that triggers permission dialogs for things like loading (1 or n) external data sets. "Uploading" files, loops that run more than 30,1000 or 10k rounds, access to camera/photos, canvas, LLM etc
You start a topic by pasting a dataset or a link to some JSON csv, xml, sql files, etc, then add a description or some other rage bate to start a conversation.
Seems much fun and a great way to start getting into coding.
It obviously requires a lot of figuring out (by better men than me) but it seems a worthwhile adventure.
The halting problem can just be figured out experimentally in production. Users might need to earn or lose privileges over time as things escalate. Admins might approve a script and allow running it without prompts or pressing the button.
Perhaps entire applications can be written this way. I much enjoyed old php code with discussions in threaded comments. (The cache will save us.)
Tried AI for a standalone html page snippet renderer[0].
I tried just feeding it snippets one by one and that worked, but the raymarching ones it could not get going until I gave it the dwitter github repos. Now most, but not all (simple to fix manually though), work.
Also interesting to see Claude is terrible at trying to write the art (the demos itself) and seeing what it tries to do; not surprising given the challenge.
I wonder what could be a criteria for demos like this that doesn’t devolve into code golf nor spawns giant files.
I’m not saying that to disparage the content mind you; it is amazing. Just wondering which potential rules could make similar results and “normal”/readable code, as that would be awesome to see.
Thanks for sharing, I'm into 8 bit retro games lately and I assumed this pico8 was some playdate emulator for some reason. My ignorance was rectified today.
The compression tricks used in standalone Javascript demos are significantly more cursed. The meta is to concatenate a compressed binary payload and a HTML/JS decompression stub in the same file, abusing the fact that HTML5 parsers are required to be absurdly tolerant of malformed documents. Nowadays it's done using raw DEFLATE and DecompressionStream, but before that was available they would pack the payload into the pixels of a PNG and use Canvas to extract the data.
What a cool concept to turn the SM 140 chars concept into a programming demo list! Is there a reverse to this that work to compress a JS script into 140 chars to help show some creative ideas without needing to fully understand the code? Also, how good is AI at generating these in your experiences?
Apologies for the quick reboot of the servers. I learned from last time this happened and resized the digitalocean droplet.
I suppose you could make the same argument about size-based competitions at demo parties; the limit is strictly 4096 bytes but everyone + their dog is using Crinkler to compress a 12-20KB executable down to that size. In fact, part of the effort is in aligning data and tweaking constants with Crinkler's algorithm in mind, to make the raw data more amenable to compression! But at least then, it's not a constant compression (turning any 194 ASCII characters into 97 Unicode characters which with embedded decoder makes exactly 140 characters)
Fractals: https://www.dwitter.net/h/fractal
Dynamical systems / chaos: https://www.dwitter.net/h/chaos
Strange attractors: https://www.dwitter.net/h/attractor
Spiral-based generative art: https://www.dwitter.net/h/spiral
Fireworks simulation: https://www.dwitter.net/h/fireworks
Procedural scenes: https://www.dwitter.net/h/scene
Wavelets (3D and 2D): https://www.dwitter.net/h/wavelet
And the most popular dweets: https://www.dwitter.net/top/all
e.g. https://github.com/developit/htm and https://lit.dev/docs/components/rendering/
https://gisthost.github.io/?8d537c4a3b2331e7c6c45d31f1900330
The Trashcan in that sampler is stored as
decoded to a SVG D path by which can be turned into an image by The Sampler contains all of the images in the html file and still comes out to be only 13k. Probably quite a bit less if you removed the ray tracer and scenes.sql`SELECT * FROM users WHERE id = ${userId}`
const q = gql` query GetUser { user(id: ${userId}) { name email } } `;
[1] https://youtu.be/JsAetmgJRss?si=AxIFySX7ktzu5GL5&t=193
[2] https://github.com/jed/140bytes/wiki/Byte-saving-techniques
[3] https://gist.github.com/jed/982883
if(!window.g){gc=document.createElement('canvas');g=gc.getContext('webgl');c.parentNode.replaceChild(gc,c);k=Object.keys(g.__proto__);d='void main(){gl_';str = `g395(g41,g318())g310(g41,new Float32Array([0,0.5,0,-0.5,-0.5,0,0.5,-0.5,0]),g46);a=g322(g139)g382(a,'attribute vec3 c;${d}Position=vec4(c,1);}')g313(a);b=g322(g138)g382(b,'${d}FragColor=vec4(1,0,1,0);}')g313(b);d=g320()g302(d,a)g302(d,b)g376(d)g393(d)g435(0,3,g128,0,0,0)g406(0)g404(g12,0,3)`.replace(/g(\d+)/g,"\ng.\${k[$1]}");eval("eval(`"+str+"`)")}
I assume there are far more advanced methods than what I managed to do, so it's probably possible to actually get that to 140 characters.
x.arc&&c.parentNode.replaceChild(g=c.cloneNode(),c);x=g.getContext`webgl`;f=Float32Array;eval(`d=320()395(41,318())310(41,new f([0,n=0.5,-n,-n,n,-n]),46)${z='382(a=322(13'}9),'attribute vec2 c;${d='void main(){gl_'}Position=vec4(c${w=",0,1);}')313(a)302(d,a)"}${z}8),'${d}FragColor=vec4(0,1${w}376(d)393(d)435(0,2,128,0,0,0)406(0)404(12,0,3)`.replace(/\d{2,}/g,"\nx[Object.keys(x.__proto__)[$&]]"))
Clever math does the heavy lifting on dwitter.net
Dwitter used the definition of what Twitter at the time used. The solution of https://beta.dwitter.net/ mitigates the lack of accessibility of the encoding, while keeping the fixed target at the same time introducing a new practical target.
The Math.sin css color encoder were concessions to practicality, there has been much discussion as to whether a Dwitter 2 should simply include additional characters required to the same preconfig and let people do it themselves, and let them use the characters for something else if they can think of something.
I'd like something like this just to see how creative people get in doing that. How small can you make a bit of JavaScript that copies every static method of Math to window?
and creativity is the goal here. No matter how people stretch the system, it can't be denied they have been doing so creatively.
You start a topic by pasting a dataset or a link to some JSON csv, xml, sql files, etc, then add a description or some other rage bate to start a conversation.
Seems much fun and a great way to start getting into coding.
The halting problem can just be figured out experimentally in production. Users might need to earn or lose privileges over time as things escalate. Admins might approve a script and allow running it without prompts or pressing the button.
Perhaps entire applications can be written this way. I much enjoyed old php code with discussions in threaded comments. (The cache will save us.)
Reminds me of Richard
https://stackoverflow.com/questions/184618/what-is-the-best-...
Show HN: A Stargate in 140 chars of JavaScript - https://news.ycombinator.com/item?id=25088683 - Nov 2020 (55 comments)
Dwitter – A social network for short JavaScript demos - https://news.ycombinator.com/item?id=13700698 - Feb 2017 (71 comments)
Tried AI for a standalone html page snippet renderer[0].
I tried just feeding it snippets one by one and that worked, but the raymarching ones it could not get going until I gave it the dwitter github repos. Now most, but not all (simple to fix manually though), work.
Also interesting to see Claude is terrible at trying to write the art (the demos itself) and seeing what it tries to do; not surprising given the challenge.
[0] https://claude.ai/public/artifacts/d8b357df-5982-48c6-be58-7...
https://xcancel.com/search?q=%23%E3%81%A4%E3%81%B6%E3%82%84%...
It's the weekend! Take a journey back in time...
https://csszengarden.com/
I’m not saying that to disparage the content mind you; it is amazing. Just wondering which potential rules could make similar results and “normal”/readable code, as that would be awesome to see.
https://www.shadertoy.com/
https://x.com/search?q=%23postcart%20%23pico8
maybe #tweetcart is a better hash tag
https://x.com/search?q=%23tweetcart
Try saving this Normal HTML File and open it in a text/hex editor: https://0b5vr.com/domain/domain.html