Then I realized that nothings stopping me from writing software how I want and feel is best. I've stopped using LLMs completely and couldn't be happier. I'm not even struggling at work or feeling like I'm behind. I work on a number of personal projects too, all without LLMs, and I couldn't feel better.
Blocking “LLM training” in a license feels satisfying, but I’ve run into three practical issues while benchmarking models for clients:
1. Auditability — You can grep for GPL strings; you can’t grep a trillion-token corpus to prove your repo wasn’t in it. Enforcement ends up resting on whistle-blowers, not license text.
2. Community hard-forks — “No-AI” clauses split the ecosystem. Half the modern Python stack depends on MIT/BSD; if even 5 % flips to an LLM-ban variant, reproducible builds become a nightmare.
3. Misaligned incentives — Training is no longer the expensive part. At today’s prices a single 70 B checkpoint costs about \$60 k to fine-tune, but running inference at scale can exceed that each day. A license that focuses on training ignores the bigger capture point.
A model company that actually wants to give back can do so via attribution, upstream fixes, and funding small maintainers (things AGPL/SSPL rarely compel). Until we can fingerprint data provenance, social pressure—or carrot contracts like RAIL terms—may move the needle more than another GPL fork.
Happy to be proven wrong; I’d love to see a case where a “no-LLM” clause was enforced and led to meaningful contributions rather than a silent ignore.
MIT is a donation of your labour to corporations. With a stronger license, at least they're more likely to contribute back or to pay you for a looser license.
If you want a permissive license MIT is perfectly reasonable. If you want more restrictions or stronger copy-left then don't pick MIT.
No, MIT is a donation of your labor to the public. That includes corporations, yes, but it is not only corporations.
If you set up a login flow with a click through that explicitly sets the terms of access, specifying no cost for access by a person, and some large cost for access by AI.
Stepping past this prompt into the content would require an AI to either lie, committing both fraud and unauthorized access of content.. or behave truthfully, opting in the proprietor of the API to the associated costs.
In either case, the site operator can then go after the company doing the scraping to collect the fees as specified in the copyright contract (and perhaps some additional delta of punitive fines if content was accessed fraudulently).
Unless you are willing to spend yourself into financial ruin pursuing legal action against some faceless megacorp - it literally doesn't matter what license you use.
I've lived enough to know there is "what should be" and then there is what actually happens in reality. We don't live in a reality where everyone just does things out of the goodness of their heart...
Adding some text to your project, hosted on a public website for all to see means some people will take your code regardless of the license or your intent - and, realistically, what are you going to do about it? Nothing...
So... please, let's get off this GPL high-horse. It's not some end-all-be-all holy text that solves all of the world's problems.
@HACKERALERT Your decision may be somewhat irresponsible towards those who donated to the audit.
Surely a recent audit only increases the odds of someone assuming responsibility for a fork. Knowing there is a solid baseline to proceed from.
Perhaps many would have refused to donate if they knew that the project would be archived in a year. Collecting for audit and then archiving the project is, in a way, a violation of expectations.
The more and more you start modifying code after the audit, the more and more useless the audit becomes.
I don't think the author had explicit plans to do this a year ago.
(I'm not trying to throw shades at the author. I know they have no obligation to maintain an open source project. I'm simply having a hard time gasping what's happening.)
I don't get it either, because that has always been the case, thus most of his post is borderline non sense.
Imo, what happened is that he took the opportunity of him entering academia to vent some stuff and quit maintaining his project.
Yes, making software development cheaper has been the main priority of the industry for a long time. The new development is that there's now a magic "do what I want" button that obviously won't quite do what you want but it's so cheap (for corporations, not humanity...) that you might as well pretend it does. (Compared to paying professionals who might even care about doing a good job, that is.)
He explained the reasons he went into the academia, which is because of the AI, and AI is not the reason he stopped with development.
- The author enjoyed writing quality open source code
- The author needs to make strategic decisions for his own career and livelihood and he doesn't have enough bandwidth for both
I don't feel he is happy about the decision he needs to make and he is pointing to something dark happening to software development and open source.
Now this is not new, and didn't start with LLM. I am sure if we ask the OpenBSD devs what they think about the modern mainstream open source ecosystem, docker, full stack development, etc. they see it like we might look at LLM generated code. This is just a question how much of a purist you are.
Could this mean that he has been approached by some "men in black" asking to insert some backdoor in the code or to stop working on it, together with a gag order ? (actually I also wondered the same a long time ago with Truecrypt, even though to my knowledge no backdoor has ever been found in further audits...)
My projects, my rules.
It begins with a prompt directed at Gemini, followed by what appears to be an AI-generated response. Are these actual AI responses or is the developer writing their parting message in a whimsical way? I'm genuinely confused. Help much appreciated!
I'm unsure if the post is actually created by Gemini or the developer's imitation. I suspect the latter.
The simpler the software the less urgent the concern but "I haven't had a problem in the last 2 years" is something I could say of most software which I end up needing to update, and it makes sense to make myself able to do so at a convenient time rather than the moment the problem occurs.
This project seems popular enough I'm sure eaiting a bit and seeing who the successor project is would be a safe bet as well though.
What worries me more (on coding related impact of AI - because of course all the impact on fake news and democracies are much more worrying IMO) is having to deal with code written by others using AI (yes, people can write shity code on their own, but with manageable pace)
But oh boy have I seen a lot of mediocre coders get away with mediocrity for a long time — there's a big risk that employers won't care about the quality, just the price, for long enough that the developments in AI are no longer foreseeable.
Yes, fake news driving by AI slop is a big problem, but that is only enabled by social media personality fiefdoms.
The shit is going to hit the fan if 10% of the highest paid US working population is laid off for AI outsourcing. That kind of social change brings revolution. and thats before the fracture of US social fabric.
Then why isn't the software job market recovering
Not only that, but it's pushing market rates down significantly. I'm making about 60% of what I made the past few years... I could only handle not having income for so long. I was juggling two jobs for a while, but just couldn't manage it. Hoping to pick up some side work to fill the gaps. Have to face it, a lot of the high pay contract software jobs have just dried up for now.
Daniel Stenberg (from the curl project) has blogged a bunch about AI slop seeping into vulnerability reports[1], and if the same happens to code contributions, documentation and so forth that can help turning a fun hobby project into something you dread maintaing.
[1] https://daniel.haxx.se/blog/2025/07/14/death-by-a-thousand-s...
Lots of graphic designers lost their jobs or at least a lot of their work now that image generation models have gotten decent at rendering text. Now any idiot can whip up some advertising graphics at half the quality of a designer, but in 1/10th the time and 1/100th the cost (or even for free!). It doesn't matter that it looks like ass and makes no sense in context, they produced an acceptable result for a fraction of the cost.
Quality does not matter in the market, it never has. Whoever can produce the most slop at the lowest price nearly always wins. Yes, there are exceptions, many of them even. But not enough to employ nearly as many of us as there are now.
This to me is the crux of the whole thing.
Almost like a knitter throwing away their needles because they saw a loom.
What your grasping for is https://en.wikipedia.org/wiki/Power-loom_riots
Where a precipitous drop in earning power, combined with longer working hours, high inflation and large companies making people unemployed cause large social unrest.
And yeah, I can see why they rioted.
* I’m not passionate about it anymore
* I’m tired
* I want to repurpose my free time
* I’m not adding enough value compared to other options now available
In the end, it’s pointless to argue about why someone feels the way they do. If they are firm on their stance, don’t waste anybody’s time, no matter how irrational their argument is. Give up trying to be right.
Probably this guy should have just stopped engaging directly with some of the dialogue, but the fact that he is exploring the idea of trying to hand it off in some manner tells me he really does care about the project.
Software engineering pays because companies want people to develop software. It pays so well because it's hard, but the coding portion is become easier. Vibe coding and AI is here to stay, the author can choose to ignore it and go preach to a dying field (specifically, writing code, not CS), or embrace it. We should be happy we no longer need to type away if and for loops 20 times and instead can focus on high level architecture.
those are still faster and cheaper and more predictable than an LLM in many cases
ImGUI and Nuklear each have 20+ backends in their repos: <https://github.com/ocornut/imgui/tree/master/backends> <https://github.com/Immediate-Mode-UI/Nuklear/tree/master/dem...>
I'm holding out hope that there will always be boutique/edge software to be written, which requires enough design and care to be mentally challenging and engaging - the craftsmanship kind.
When AlphaGo was announced, I had to keep telling people that "It's not like computers win at Go, it's just that we now have a tool that makes us way better at Go". If an alien race showed up and challenged us to Go to save the species, would we use a Go player or a AlphaGo, if we had to choose?
The problem is LLMs aren't like that, because software isn't like Go. And, they really are annoying to use, frustrating to redirect all the time, and generally cannot do what you want precisely, without putting in more mental energy to provide context and decompose further than you'd need to do the damn thing yourself. And then you lose an hour/two of flow, which is the reward for the whole process.
But at certain times they are a godsend and they have completely replaced some of the more boring parts of my job. I wouldn't want to lose them, not at all.
Like the author, I don't think we're heading to a healthy balance where LLMs help us be better at our job. I do think the hype is going in the wrong direction, and I do worry for the state of our field (Or at least the _ideals_ of our field). Call me naive, but I also thought it mattered what the code was.
Software engineering provides a window into reality in a way which exposes you to its full complexity. It changes your brain, you start seeing complexity everywhere around you. You start seeing problems that nobody else sees. This is why I got into coding... But now the industry often feels like it's leading you astray and preventing you from truly flourishing.
This is sad because being skilled at coding feels great and it shapes your reality in a very positive way. Being able to think clearly is a great gift and a worthy goal to strive for. Having a logical mind feels inherently good. Being able to approach any topic, with anyone and maintaining full logical consistency feels good.
That was strangely...something. Simultaneously not what I expected and yet just nailing the vibe of vibe slop frustration.
Large and centered on the authors website: https://evansu.com/
(This short comment exchange between us is also a meta commentary on that. Because our comments are much shorter than all the AI summaries, and at the same time, we add nuance and clarification to the ideas exposed, something the AI summaries don't do.)
yeah but _what_ part of AI research. There are loads of it that are nothing to do with slop, and might even have practical and useful benefits....
Replace "software engineering" with literally anything and this is a true statement since the dawn of civilization.
IMO anyone who understands AI at a technical level will understand that this won't happen. No matter how many parameters, training and compute you throw at it, putting AI in direct charge of anything that's critical and not entirely predictable is going to backfire. Though, based on response from this author, it should be apparent that his response comes from a place of emotion, misunderstanding, and likely conformism to dogmatic anti-AI rhetoric of the same nature, rather than actual reason and logic.
But in my opinion, it's a bit hypocritical to blame / be mad at LLMs etc. ruining the fun of coding because & then use AI generated profile pictures.
Why not draw your ghibli styled profile picture yourself? Why use an AI generated image? Doesn't using image generators ruin the fun of drawing? Vibe-drawing?
> He criticizes "Large tech companies and investors" for prioritizing "vibe coding," but not a specific company or individual.
You could rewrite this generated response to match artists point of view as well
lol
As an example:
> you're not helping anything or making some resonant statement with that thing above or your avatar
> Sorry for breathing and producing CO2.
That's not what the commenter argued, and that response is incredibly petty. It's a way to defuse the argument entirely (is that a straw-man? no idea).
This LLM-fuelled rant/departure is a thought-provoking expression of frustration from someone who focused on the 90%, not the 10% -- namely someone willing to handcraft software like an artisan.
I think we're in the mass-production era for code and nobody wants a hand-crafted anything anymore. Automated and mass-produced please. "Quantity has a quality all its own"
But even if that's true, the 1000x is going to go to far fewer humans. Maybe you're in the lucky % saved, but a lot of people won't be.
It's interesting to consider. I don't have any takes one way or the other, I'm just observing. I have no idea how all of this works out.
Is the 90% just hand crafting? I don't believe there can be no place for hand crafting things at all, because there is not any other business or human endeavor in history which has been 100% mass produced without any place for artisans in the market.
Even Automobile production has its place in the market for craftsmanship.
Really looking forwards to the 10000% pay raise
Not all people like IKEA furniture.
>Quantity has a quality all its own
You understand the meaning of this Stalin attributed quote?
I wouldn’t want that kind of quality in planes, cars, surgical robots, power plants etc.
I think it’s a good time to introduce fines for severe damages by software bugs.
> new career path is LLM research
what?? how do you square that circle. deciding to go contribute to the very thing that steered your path off course
QFT
Yes, the AI hype is real, and yes there's a desire to cut costs by using AI within companies. However, I think the maintainer (Evan Su) has a bit of a narrow view on this matter. Evan is still a student in university.
This doesn't mean his perspective or opinion should be disregarded, it's more just I think he's declaring quite a career defining absolute for himself before really having a solid foot in the industry. Frankly, this rant seems kind of fueled by intense doom-scrolling on linkedin rather than by first hand experience.
It's a very very common problem in the Demoscene, demoscene authors striving to impress in the 64k category created better generic EXE compressors to shrink more content into their 64k binaries, just when any tool is publicly released it is almost immediately flagged by AV software since virus makers scour our forums knowing that once in a while a useful tool comes to hide their malicious code.
I envision a future when people can't deduct anything by themselves and will rely on automated, flawed systems.
The problem is not knowing why and how the systems are flawed, and therefore being unable to have nuanced and truly accurate decisions.
I’m not at a computer where I could try that hypothesis right now, but back then my conclusion after testing various executables was that any unsigned Go binary would be flagged as a malware.
I didn't know about picocrypt but I already have two options for safely encrypting files: 7zip with its AES-256 (simple) and veracrypt with various algorithms (more involved but allows you to mount the encrypted vols). Actually these are already mentioned in the tool's readmy, great work: https://github.com/Picocrypt/Picocrypt?tab=readme-ov-file#co...