> I believe a sufficiently interested emacs developer with a
$100 plan and a long weekend could hammer out a pretty good first cut.
That no one's yet done it is clear evidence that no one really wanted
these things so much as they wanted to talk about it.
The disappearance of bikeshedding doesn't look like an obvious one but it's a huge loss. Imagine if every scientist resolved to refer to already written books and carry on their research in closed labs with no discussion with the rest of their field.
“AI has turned Stallman's goal of software
freedom on its head. Having your computer do whatever your
imagination can think up -- that's real freedom. The only catch is
you'll have no idea how it works. No one does.”
"Being dependent on a small number of AI vendors, who also penetrate every aspect of your life to control, report, and spy upon you, forced to use AI at work or be laid off, or just be made redundant and let off anyway, while drowing in slop from every corner of social media, news, youtube, search results, and any other cultural production -- that's real freedom!"
Other ones being to keep paying the AI bros in perpetuity, and having all your prompts, data, and IP getting stolen by the same people with the goal of eventually undermining your own ability to think. We will all be as free as the people living in Pods in Matrix.
I'm also not sure how that ended up the definition of real freedom. Maybe some Spinoza is in order but genuine freedom is proper understanding of one's own, and others, nature and causes. That is to say one is free to the extent that one has adequate knowledge.
To live in ignorance, that is to be pushed, pulled or compelled by what you don't understand, for example because you've outsourced your life to the magical mystery programming box, is in his terminology bondage.
Which also explains why all these AI psychosis people sound like crack addicts. Stallman is still completely correct about what software freedom is
⟩ Refactor these components into standalone modules that communicate across well-defined APIs. Write the API specifications to docs/API.md. Add a note to AGENTS.md instructing agents to preserve the modular seams and focus work on one module at a time.
I more or less agree with the article but we can move it one step further. Nowadays we don't even need text editors no matter how advanced they are. I am a VS Code user and I have not opened VS Code in more than 3 months because I don't need it anymore. Opening a text editor and editing code by hand seems like so ... yesterday.
This definitely is a new era and, personally, I love it! (Also I have broken my right arm so I can't write code even if I wanted to. I'm dictating this text.)
I was saying almost this same thing to a coworker. I'm building something that's like Emacs, but for agents, and it's mostly a Markdown-forward tool for managing all the agents and agent coding harnesses, and it's a vendor-agnostic things....If anything it's like a newer, faster, fancier org-mode, but knows how to inject all the right context into all the agentic projects.
This, after configuring Emacs to talk to LLMs. But that's not the right angle now for me....
> RMS appeared in a recent video recounting his CSAIL work on a circuit analyzer in which he, and I'm paraphrasing, "didn't model the circuit as a system of equations to be solved, but rather approached it much as a human would" by applying a series of diagnostics.
Like demanding commented machine code between 1985 and 2025.
Sure you can get it, but that's not where the software lives even if it's how the software runs.
Reverse engineering closed source software is the simplest thing in the world now. The only reason why more of it isn't done is because it's so boring even minding the AI isn't worth most people's time.
"We may have felt safe bringing a gun to a knife fight, but now everyone's just sending a drone." Perfectly said. For those of us concerned about editor efficiency, it's hard to know where to go from here. For years I'd been working on a design tool that was inspired by Vim - entirely keyboard driven using mnemonic keychords. But LLMs really took the wind out of my sails, because eh, why bother now? Somewhere deep down I feel that it's still viable, because I want it to be true that perfectly deterministic tooling is still important. I just don't know if it will be considered important by everyone else.
> "I want it to be true that perfectly deterministic tooling is still important."
Of course it is. The AI coded apps only (debatably) work as end-user apps. They are thin little bumps on top of the software mountain. So much, including all the AI tools, depends on efficient, perfectly deterministic tooling.
It is almost hypocritical for AI enthusiasts to say we don't need to look at code anymore, while everything they do is build on piles and piles of infrastructure and tools that are carefully hard-coded, deterministic building blocks. The whole thing would come crashing down if everyone embraced vibe coding at every level.
I am just constantly baffled by people thinking that because we can make (crappy) end-user apps and one-off tools, no one needs to code anymore. Does any build solid tools and libraries? Does anyone still get pleasure from building a rock solid, snappy app?
I can feel all the AI coded apps and website turning to ever buggier sludge under my fingertips, and people are rejoicing that this is the way of the future.
I understand the sentiment. I do think it's important to experience it for yourself - not to convince you (I'm less and less convinced as the months go by) - but just to understand the visceral feeling that leads people into AI psychosis.
At times it feels literally magical. But more and more I feel like my brain has slumped onto the couch, lazily barking at a butler to do my bidding. Like a weird kind of professional brainrot.
I read about a guy who was doing a documentary about crack addicts. They told him he needed to try it once to understand. He did. He then got addicted, lost everything and spent years on the street. Years later he finally got sober and finished the documentary.
>Of course it is. The AI coded apps only (debatably) work as end-user apps. They are thin little bumps on top of the software mountain. So much, including all the AI tools, depends on efficient, perfectly deterministic tooling.
Most of that is also written via AI these days... Even the Linux kernel got that bug.
> But there's one glaring headscratcher, and that's its full-blown,
self-contained implementation of a terminal emulator, in Rust no less. In the before times, a developer would undoubtedly seek such a heavy
lift from off-the-shelf kit, but the holmes replicated over a
coffee break what a billionaire with infinite resources had been
toiling over the last five years.
Is this really true, though? Did model improvement since Ghostty began simply eclipse Ghostty and anything achievable with that now "legacy" code base? Would that project be better off abandoned or started over from scratch with Opus 5.5? Or is there something more there?
> sealed up transistors can reproduce the same outputs given the same inputs
with unerring precision. The power to unambiguously differentiate an
on- versus an off-signal -- the binary "digital" -- might seem
trivial, but is 100% the reason why we're so comfortable handing over
the reins to computers.
Is the author not familiar with the ECC memory business?
Here's a fun question - if AI is trained on OSS, particularly with licenses that are copy left... wouldn't that make all the code it generates copy left? wouldn't that mean the weights themselves should all be open sourced.
> The irony is rich that AI has turned Stallman's goal of software freedom on its head. Having your computer do whatever your imagination can think up -- that's real freedom. The only catch is
you'll have no idea how it works. No one does.
Why are people so sure this will keep scaling? We're still so far away from me being able to get anything even remotely close to what I can imagine using a free model running on hardware I own.
Then you need to remember there are a lot of very powerful people currently invested in not letting us have any of this, so I really don't think it's time to call free software irrelevant yet.
>ut at the time, rules-based
programming was still novel enough to be considered a viable approach
to AI.
Pick MLite from https://t3x.org and you'll get it pretty fast. Or with invariants under Lisp. Or its weird cousing, CLIPS, born from Common Lisp clauses.
On code quality, Neomacs (an Emacs clone in Rust, it has Elisp support) crashes a lot. And by a lot, I mean once you hit the menu with the mouse, it segfaults.
>he only catch is
you'll have no idea how it works. No one does.
Is not about the rules, but the relations. It's a network of clauses, and not something akin to if-then-else with flow controlling. There's no actual "flow".
It's a highly experimental thing someone decided to mess around with, without any prior design discussion with any Emacs core devs. It's like a forty-minute-old baby and you're asking it to parallel park without crashing. What's your point here, I don't get it?
GNU/Emacs is incredibly stable. It's one of the most reliable pieces I've ever had; I sometimes run it for months without ever restarting. Which I can't say about pretty much any app on my system. Including the system itself.
Fable 5.2 is designing entire devices now, from PCB layout to case assembly.
If you're in a STEM program at university, quit. By the time you graduate, you will have negative value proposition compared to Claude. Get into business. The only way you're going to make it is to join the ownership class. Or pick up a torch or pitchfork and join the revolution.
This is such a wrong way of looking at it. The fact that AI is getting better and giving us better toolings is more of a reason to go and get an education.
The disappearance of bikeshedding doesn't look like an obvious one but it's a huge loss. Imagine if every scientist resolved to refer to already written books and carry on their research in closed labs with no discussion with the rest of their field.
To live in ignorance, that is to be pushed, pulled or compelled by what you don't understand, for example because you've outsourced your life to the magical mystery programming box, is in his terminology bondage.
Which also explains why all these AI psychosis people sound like crack addicts. Stallman is still completely correct about what software freedom is
We need to be thinking about modularity with well-defined APIs now, to isolate the chunks created by AI coding. Maintainabilty requires firewalls.
⟩ Refactor these components into standalone modules that communicate across well-defined APIs. Write the API specifications to docs/API.md. Add a note to AGENTS.md instructing agents to preserve the modular seams and focus work on one module at a time.
This definitely is a new era and, personally, I love it! (Also I have broken my right arm so I can't write code even if I wanted to. I'm dictating this text.)
This, after configuring Emacs to talk to LLMs. But that's not the right angle now for me....
bitter-lesson-pilled RMS before GTA VI
Like demanding commented machine code between 1985 and 2025.
Sure you can get it, but that's not where the software lives even if it's how the software runs.
Reverse engineering closed source software is the simplest thing in the world now. The only reason why more of it isn't done is because it's so boring even minding the AI isn't worth most people's time.
Of course it is. The AI coded apps only (debatably) work as end-user apps. They are thin little bumps on top of the software mountain. So much, including all the AI tools, depends on efficient, perfectly deterministic tooling.
It is almost hypocritical for AI enthusiasts to say we don't need to look at code anymore, while everything they do is build on piles and piles of infrastructure and tools that are carefully hard-coded, deterministic building blocks. The whole thing would come crashing down if everyone embraced vibe coding at every level.
I am just constantly baffled by people thinking that because we can make (crappy) end-user apps and one-off tools, no one needs to code anymore. Does any build solid tools and libraries? Does anyone still get pleasure from building a rock solid, snappy app?
I can feel all the AI coded apps and website turning to ever buggier sludge under my fingertips, and people are rejoicing that this is the way of the future.
I refuse. I will not touch Claude.
At times it feels literally magical. But more and more I feel like my brain has slumped onto the couch, lazily barking at a butler to do my bidding. Like a weird kind of professional brainrot.
Most of that is also written via AI these days... Even the Linux kernel got that bug.
> But there's one glaring headscratcher, and that's its full-blown, self-contained implementation of a terminal emulator, in Rust no less. In the before times, a developer would undoubtedly seek such a heavy lift from off-the-shelf kit, but the holmes replicated over a coffee break what a billionaire with infinite resources had been toiling over the last five years.
Is this really true, though? Did model improvement since Ghostty began simply eclipse Ghostty and anything achievable with that now "legacy" code base? Would that project be better off abandoned or started over from scratch with Opus 5.5? Or is there something more there?
These are honest questions.
Is the author not familiar with the ECC memory business?
So, no.
Why are people so sure this will keep scaling? We're still so far away from me being able to get anything even remotely close to what I can imagine using a free model running on hardware I own.
Then you need to remember there are a lot of very powerful people currently invested in not letting us have any of this, so I really don't think it's time to call free software irrelevant yet.
The AI tooling seems too unstable still.
Pick MLite from https://t3x.org and you'll get it pretty fast. Or with invariants under Lisp. Or its weird cousing, CLIPS, born from Common Lisp clauses.
On code quality, Neomacs (an Emacs clone in Rust, it has Elisp support) crashes a lot. And by a lot, I mean once you hit the menu with the mouse, it segfaults.
>he only catch is you'll have no idea how it works. No one does.
Is not about the rules, but the relations. It's a network of clauses, and not something akin to if-then-else with flow controlling. There's no actual "flow".
It's a highly experimental thing someone decided to mess around with, without any prior design discussion with any Emacs core devs. It's like a forty-minute-old baby and you're asking it to parallel park without crashing. What's your point here, I don't get it?
GNU/Emacs is incredibly stable. It's one of the most reliable pieces I've ever had; I sometimes run it for months without ever restarting. Which I can't say about pretty much any app on my system. Including the system itself.
If you're in a STEM program at university, quit. By the time you graduate, you will have negative value proposition compared to Claude. Get into business. The only way you're going to make it is to join the ownership class. Or pick up a torch or pitchfork and join the revolution.
...which is rapidly running away from you. Ignore this nonsense and do useful work. It will be okay.
I hope you are not advising any young people out there.