I am thinking of that quotation that said [paraphrasing] "90% of my skills went to $0, but the other 10% are now worth 1000x"
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"
LLMs are glorified "LMGTFY" tools. AI assistance doesn't make people experts at anything. Some genz vibe coder isn't getting your job guys calm the heck down.
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.
I like the creativity behind this. And I feel sorry for them that the current wave of AI has lead to them abandoning their pet project. Maybe they will pick up the project again, once the dust has settled. At the end, at least for me, they are pet projects for exactly that reason: An escape route from the day to day business. One should still be proud of what they achieved in their spare time. I don’t care if my job requires me to use K8s, Claude or Docker. Or if that’s considered "industry standard".
(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.)
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
As a complete outsider looking at this, without additional context, I just have a hard time believing there aren’t more reasons, they’re just not willing to share them:
* 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.
I've felt similar to the author, a sort of despair that the only point of writing software now is to prop up the valuation of AI companies, that quality no longer matters, etc.
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.
This is also a good opportunity to remember that MIT is not a strong enough open source license, and if you want to prevent corporations making money off your work, make it AGPL or even SSPL, plus a statement that AI training creates a derivative work (the latter may or may not have any legal effect).
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.
MIT isn’t “weak” because it allows LLM training; it’s weak because it puts zero obligations on the recipient.
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.
Immediate mode UI toolkits are designed for pluggable backends, some even can discover the appropriate backend at runtime. If you're writing a game, you're expected to (and actually, you must) build your own integration.
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.
This was really huge. I actually had to pass it to an LLM to get an abstract ....
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...
Damn. The part about quality over quantity really hit home. I also got into software engineering as a way to exert craftsmanship and am disappointed with the state of things. It's potentially a great field for people who are interested in the pursuit of perfection. There are few areas which are as complex and take so much time and effort to master.
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.
TLDR: Author of an open-source project has a crashout over other people using LLMs for coding, believes that AI will replace all developers, and decides to preemptively give up on software engineering entirely because of that.
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.
I understand the author's sentiment but industries don't exist solely because somebody wants them to. I mean, sure, hobbies can exist, but you won't be paid well (or even at all) to work with them.
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.
In regards to "As long as you can run the code, archiving this project means nothing, really." I think this section misses the main concern of archived software - what happens when one of those bugs is run into (either something not yet noticed or something due to external changes down the road) and there is no actively maintained version (which could include one you're willing to hack st yourself) to just update to?
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.
I'm starting to feel kinda old and out of the loop. Could someone please explain the conversational style of this post?
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!
35 comments
[ 2.5 ms ] story [ 51.3 ms ] threadThat was strangely...something. Simultaneously not what I expected and yet just nailing the vibe of vibe slop frustration.
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"
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.
My projects, my rules.
(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.)
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
QFT
This to me is the crux of the whole thing.
Almost like a knitter throwing away their needles because they saw a loom.
* 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.
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.
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.
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.
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.
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...
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.
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.
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.
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.
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!