The irony of a five sentence article making giant claims isn't lost on me. Don't get me wrong: I'm amenable to the idea; but, y'know, my kids wrote longer essays in 4th grade.
Guessing this isn’t going to be popular here, but he’s right. AI has some use cases, but isn’t the world-changing paradigm shift it’s marketed as. It’s becoming clear the tech is ultimately just a tool, not a precursor to AGI.
> Meanwhile, my cofounder is rewriting code we spent millions of salary on in the past by himself in a few weeks.
If the LLM generating the code introduced a bug, who will be fixing it? The founder that does not know how to code or the LLM that made the mistake first?
Key thing here. The code was already written, so rewriting it isn't exactly adding a lot of quantifiable value. If millions weren't spent in the first place, there would be no code to rewrite.
Doesn't this imply that you were not getting the level of efficiency out of your investment? It would be a little odd to say this publicly as this says more about you and your company. The question would be what your code does and if it is profitable.
In this thread: people throwing shade on tech that works, comparing it to a perfect world and making weird assumptions like no tests, no E2E or manual testing just to make a case. Hot take: most SWEs produce shit code, be it by constraints of any kind or their own abilities. LLMs do the same but cost less and can move faster. If you know how to use it, code will be fine. Code is a commodity and a lot of people will be blindsided by that in the future. If your value proposition is translating requirements into code, I feel sorry for you. The output quality of the LLM depends on the abilities of the operator. And most SWEs lack the system thinking to be good here, in my experience.
As a fractional CTO and in my decade of being co-founder/CTO I saw a lot of people and codebases and most of it is just bad. You need to compare real life codebases and outputs of developers, not what people wished it would be like. And the reality is that most of it sucks and most SWEs are bad at their jobs.
Good luck with fixing that future mess. This is such an incredibly short sighted approach to running a company and software dev that I think your cofounder is likely going to torpedo your company.
All the productivity enhancement provided by LLMs for programming is caused by circumventing the copyright restrictions of the programs on which they have been trained.
You and anyone else could have avoided spending millions for programmer salaries, had you been allowed to reuse freely any of the many existing proprietary or open-source programs that solved the same or very similar problems.
I would have no problem with everyone being able to reuse any program, without restrictions, but with these AI programming tools the rich are now permitted to ignore copyrights, while the poor remain constrained by them, as before.
The copyright for programs has caused a huge multiplication of the programming effort for many decades, with everyone rewriting again and again similar programs, in order for their employing company to own the "IP". Now LLMs are exposing what would have happened in an alternative timeline.
The LLMs have the additional advantage of fast and easy searching through a huge database of programs, but this advantage would not have been enough for a significant productivity increase over a competent programmer that would have searched the same database by traditional means, to find reusable code.
As a software engineer this scares me from an employment perspective but I also use Claude Code to produce 90% of the code I commit now (after reviewing and revision of course) so there’s that… ;)
When I read the blog post, the impression I get is that the author is referring to the proposed "business" of licensing or selling "generative AI" (i.e., making money for the licensor or seller), not whether generative AI is saving money for any particular user
The author's second reference, an article from The Atlantic, describing the copyright liability issues with "generative AI", has been submitted to HN four times in the last week
AI Memorization Research (theatlantic.com)
2 points by tagyro 5 hours ago | flag | past | discuss
AI's Memorization Crisis (theatlantic.com)
2 points by twalichiewicz 1 day ago | flag | past | 1 comment
AI's Memorization Crisis (theatlantic.com)
3 points by palad1n 4 days ago | flag | past | 1 comment
AI's Memorization Crisis (theatlantic.com)
4 points by casparvitch 4 days ago | flag | past | discuss
> I myself am saving a small fortune on design and photography and getting better results while doing it.
Tell me you have bland taste without telling me you have bland taste. But if your customers eat it up and your slop manages to stand out in sea of slop, who am I to dislike slop.
Exactly, our venture studio that partnered with our startup collapsed. The code there team wrote was that took two years was terrible and didn't fully function. My CTO and I are rewriting 60% of the code with AI! Now everything works with bugs.....
I find it a bit odd that people are acting like this stuff is an abject failure because it's not perfect yet.
Generative AI, as we know it, has only existed ~5-6 years, and it has improved substantially, and is likely to keep improving.
Yes, people have probably been deploying it in spots where it's not quite ready but it's myopic to act like it's "not going all that well" when it's pretty clear that it actually is going pretty well, just that we need to work out the kinks. New technology is always buggy for awhile, and eventually it becomes boring.
>Generative AI, as we know it, has only existed ~5-6 years
Probably less than that, practically speaking. ChatGPT's initial release date was November 2022. It's closer to 3 years, in terms of any significant amount of people using them.
I don't think LLMs are an abject failure, but I find it equally odd that so many people think that transformer-based LLMs can be incrementally improved to perfection. It seems pretty obvious to me now that we're not gonna RLHF our way out of hallucinations. We'll probably need a few more fundamental architecture breakthroughs to do that.
It's going well for coding. I just knocked out a mapping project that would have been a week+ of work (with docs and stackoverflow opened in the background) in a few hours.
And yes, I do understand the code and what is happening and did have to make a couple of adjustments manually.
I don't know that reducing coding work justifies the current valuations, but I wouldn't say it's "not going all that well".
This feels like a pretty low effort post that plays heavily to superficial reader's cognitive biases.
I work commercializing AI in some very specific use cases where it extremely valuable. Where people are being lead astray is layering generalizations: general use cases (copilots) deployed across general populations and generally not doing very well. But that's PMF stuff, not a failure of the underlying tech.
A year ago I would have agreed wholeheartedly and I was a self confessed skeptic.
Then Gemini got good (around 2.5?), like I-turned-my-head good. I started to use it every week-ish, not to write code. But more like a tool (as you would a calculator).
More recently Opus 4.5 was released and now I'm using it every day to assist in code. It is regularly helping me take tasks that would have taken 6-12 hours down to 15-30 minutes with some minor prompting and hand holding.
I've not yet reached the point where I feel letting is loose and do the entire PR for me. But it's getting there.
I'm now putting more queries into LLMs than I am into Google Search.
I'm not sure how much of that is because Google Search has worsened versus LLMs having improved, but it's still a substantial shift in my day-to-day life.
Something like finding the most appropriate sensor ICs to use for a particular use case requires so much less effort than it used to. I might have spent an entire day digging through data sheets before, and now I'll find what I need in a few minutes. It feels at least as revolutionary as when search replaced manually paging through web directories.
I feel like I'm living in a totally different world or I'm being gaslit by LLMs when I read stuff like this and other similar comments in this thread. Do you mind mentioning _what_ language / tech stack you're in? At my current job, we have a large Ruby on Rails codebase and just this week Gemini 2.5 and 3 struggled to even identify what classes inherited from another class.
I wholeheartedly agree. Shitty companies steal art and then put out shitty products that shitty people use to spam us with slop.
The same goes for code as well.
I’ve explored Claude code/antigravity/etc, found them mostly useless, tried a more interactive approach with copilot/local models/ tried less interactive “agents”/etc. it’s largely all slop.
My coworkers who claim they’re shipping at warp speed using generative AI are almost categorically our worst developers by a mile.
I’m starting to think this take is legitimately insane.
As said in the article, a conservative estimate is that Gen AI can currently do 2.5% of all jobs in the entire economy. A technology that is really only a couple of years old. This is supposed to be _disappointing_? That’s millions of jobs _today_, in a totally nascent form.
I mean I understand skepticism, I’m not exactly in love with AI myself, but the world has literally been transformed.
I believe Gary Marcus is quite well known for terrible AI predictions. He's not in any way an expert in the field. Some of his predictions from 2022 [1]
> In 2029, AI will not be able to watch a movie and tell you accurately what is going on (what I called the comprehension challenge in The New Yorker, in 2014). Who are the characters? What are their conflicts and motivations? etc.
> In 2029, AI will not be able to read a novel and reliably answer questions about plot, character, conflicts, motivations, etc. Key will be going beyond the literal text, as Davis and I explain in Rebooting AI.
> In 2029, AI will not be able to work as a competent cook in an arbitrary kitchen (extending Steve Wozniak’s cup of coffee benchmark).
> In 2029, AI will not be able to reliably construct bug-free code of more than 10,000 lines from natural language specification or by interactions with a non-expert user. [Gluing together code from existing libraries doesn’t count.]
> In 2029, AI will not be able to take arbitrary proofs from the mathematical literature written in natural language and convert them into a symbolic form suitable for symbolic verification.
Many of these have already been achieved, and it's only early 2026.
> Many of these have already been achieved, and it's only early 2026.
I'm quite sure people who made those (now laughable) predictions will tell you none of these has been achieved, because AI isn't doing this "reliably" or "bug-free."
Defending your predictions is like running an insurance company. You always win.
In my opinion, contrary to other comments here I think AI can do all of the above already except being a kitchen cook.
Just earlier today I asked it to give me a summary of a show I was watching until a particular episode in a particular season without spoiling the rest of it and it did a great job.
I don't understand how this claim can even be tested:
> In 2029, AI will not be able to read a novel and reliably answer questions about plot, character, conflicts, motivations, etc. Key will be going beyond the literal text, as Davis and I explain in Rebooting AI.
Once you are "going beyond the literal text" the standard is usefulness of your insight about the novel, not whether your insight is "right" or "wrong".
I keep reading comments that claim GenAI's positive traits, but this usually amounts to some toy PoC that very eerily mirrors work found in code bootcamps. You want an app that has logins and comments and upvotes? GenAI is going to look amazing setting up a non-relational db to your node backend.
Aye. If you've not turned a real profit with your thing, I will default to believing that you don't know what you're talking about and are probably building toys.
It's nothing to do with AI. I didn't believe "I rewrote my application in three weeks!" claims before AI, and I don't believe them now. Most people are not able to evaluate themselves, I don't see why that would have changed.
Gary Marcus (probably): "Hey this LLM isn't smarter than Einstein yet, it's not going all that well"
The goalposts keep getting pushed further and further every month. How many math and coding Olympiads and other benchmarks will LLMs need to dominate before people will actually admit that in some domains it's really quite good.
Sure, if you're a Nobel prize winner or PhD then LLMs aren't as good as you yet, but for 99% of the people in the world, LLMs are better than you at Math, Science, Coding, and every language probably except your native language, and it's probably better at you at that too...
83 comments
[ 6.0 ms ] story [ 95.9 ms ] threadI appreciate good critique but this is not it
The irony of a five sentence article making giant claims isn't lost on me. Don't get me wrong: I'm amenable to the idea; but, y'know, my kids wrote longer essays in 4th grade.
I myself am saving a small fortune on design and photography and getting better results while doing it.
If this is not all that well I can’t wait until we get to mediocre!
Why?
Im not even casting shade - I think AI is quite amazing for coding and can increase productivity and quality a lot.
But I'm curious why he's doing this.
I was expecting a language reference (we all know which one), to get more speed, safety and dare I say it "web scale" (insert meme). :)
If the LLM generating the code introduced a bug, who will be fixing it? The founder that does not know how to code or the LLM that made the mistake first?
Key thing here. The code was already written, so rewriting it isn't exactly adding a lot of quantifiable value. If millions weren't spent in the first place, there would be no code to rewrite.
As a fractional CTO and in my decade of being co-founder/CTO I saw a lot of people and codebases and most of it is just bad. You need to compare real life codebases and outputs of developers, not what people wished it would be like. And the reality is that most of it sucks and most SWEs are bad at their jobs.
You and anyone else could have avoided spending millions for programmer salaries, had you been allowed to reuse freely any of the many existing proprietary or open-source programs that solved the same or very similar problems.
I would have no problem with everyone being able to reuse any program, without restrictions, but with these AI programming tools the rich are now permitted to ignore copyrights, while the poor remain constrained by them, as before.
The copyright for programs has caused a huge multiplication of the programming effort for many decades, with everyone rewriting again and again similar programs, in order for their employing company to own the "IP". Now LLMs are exposing what would have happened in an alternative timeline.
The LLMs have the additional advantage of fast and easy searching through a huge database of programs, but this advantage would not have been enough for a significant productivity increase over a competent programmer that would have searched the same database by traditional means, to find reusable code.
You sound like complete clones of us :-)
We’ve been at it since July and have built what used to take 3-5 people that long.
To the haters: I use TDD and review every line of code, I’m not an animal.
There’s just 2 of us but some days it feels like we command an army.
Is this because you are improving your already existing design and photography skills and business ?
Or have you bootstrapped from the scratch with AI ?
Do you mind sharing or giving a hint ?
Thanks!
The author's second reference, an article from The Atlantic, describing the copyright liability issues with "generative AI", has been submitted to HN four times in the last week
AI Memorization Research (theatlantic.com)
2 points by tagyro 5 hours ago | flag | past | discuss
AI's Memorization Crisis (theatlantic.com)
2 points by twalichiewicz 1 day ago | flag | past | 1 comment
AI's Memorization Crisis (theatlantic.com)
3 points by palad1n 4 days ago | flag | past | 1 comment
AI's Memorization Crisis (theatlantic.com)
4 points by casparvitch 4 days ago | flag | past | discuss
Yay! Let's put all the artists out of business and funnel all the money to the tech industry. That's how to build a vibrant society. Yay!
Or is any of the existing platform is used as an input for the rewrite?
Tell me you have bland taste without telling me you have bland taste. But if your customers eat it up and your slop manages to stand out in sea of slop, who am I to dislike slop.
Generative AI, as we know it, has only existed ~5-6 years, and it has improved substantially, and is likely to keep improving.
Yes, people have probably been deploying it in spots where it's not quite ready but it's myopic to act like it's "not going all that well" when it's pretty clear that it actually is going pretty well, just that we need to work out the kinks. New technology is always buggy for awhile, and eventually it becomes boring.
Probably less than that, practically speaking. ChatGPT's initial release date was November 2022. It's closer to 3 years, in terms of any significant amount of people using them.
It's a business, but it won't be the thing the first movers thought it was.
And yes, I do understand the code and what is happening and did have to make a couple of adjustments manually.
I don't know that reducing coding work justifies the current valuations, but I wouldn't say it's "not going all that well".
I work commercializing AI in some very specific use cases where it extremely valuable. Where people are being lead astray is layering generalizations: general use cases (copilots) deployed across general populations and generally not doing very well. But that's PMF stuff, not a failure of the underlying tech.
Then Gemini got good (around 2.5?), like I-turned-my-head good. I started to use it every week-ish, not to write code. But more like a tool (as you would a calculator).
More recently Opus 4.5 was released and now I'm using it every day to assist in code. It is regularly helping me take tasks that would have taken 6-12 hours down to 15-30 minutes with some minor prompting and hand holding.
I've not yet reached the point where I feel letting is loose and do the entire PR for me. But it's getting there.
I'm not sure how much of that is because Google Search has worsened versus LLMs having improved, but it's still a substantial shift in my day-to-day life.
Something like finding the most appropriate sensor ICs to use for a particular use case requires so much less effort than it used to. I might have spent an entire day digging through data sheets before, and now I'll find what I need in a few minutes. It feels at least as revolutionary as when search replaced manually paging through web directories.
The same goes for code as well.
I’ve explored Claude code/antigravity/etc, found them mostly useless, tried a more interactive approach with copilot/local models/ tried less interactive “agents”/etc. it’s largely all slop.
My coworkers who claim they’re shipping at warp speed using generative AI are almost categorically our worst developers by a mile.
I hate generative AI, but its inarguable what we have now would have been considered pure magic 5 years ago.
As said in the article, a conservative estimate is that Gen AI can currently do 2.5% of all jobs in the entire economy. A technology that is really only a couple of years old. This is supposed to be _disappointing_? That’s millions of jobs _today_, in a totally nascent form.
I mean I understand skepticism, I’m not exactly in love with AI myself, but the world has literally been transformed.
> In 2029, AI will not be able to watch a movie and tell you accurately what is going on (what I called the comprehension challenge in The New Yorker, in 2014). Who are the characters? What are their conflicts and motivations? etc.
> In 2029, AI will not be able to read a novel and reliably answer questions about plot, character, conflicts, motivations, etc. Key will be going beyond the literal text, as Davis and I explain in Rebooting AI.
> In 2029, AI will not be able to work as a competent cook in an arbitrary kitchen (extending Steve Wozniak’s cup of coffee benchmark).
> In 2029, AI will not be able to reliably construct bug-free code of more than 10,000 lines from natural language specification or by interactions with a non-expert user. [Gluing together code from existing libraries doesn’t count.]
> In 2029, AI will not be able to take arbitrary proofs from the mathematical literature written in natural language and convert them into a symbolic form suitable for symbolic verification.
Many of these have already been achieved, and it's only early 2026.
[1]https://garymarcus.substack.com/p/dear-elon-musk-here-are-fi...
I'm quite sure people who made those (now laughable) predictions will tell you none of these has been achieved, because AI isn't doing this "reliably" or "bug-free."
Defending your predictions is like running an insurance company. You always win.
Just earlier today I asked it to give me a summary of a show I was watching until a particular episode in a particular season without spoiling the rest of it and it did a great job.
> In 2029, AI will not be able to read a novel and reliably answer questions about plot, character, conflicts, motivations, etc. Key will be going beyond the literal text, as Davis and I explain in Rebooting AI.
Once you are "going beyond the literal text" the standard is usefulness of your insight about the novel, not whether your insight is "right" or "wrong".
Second of all, GenAI is going well or not depending on how we frame it.
In terms of saving time, money and effort when coding, writing, analysing, researching, etc. It’s extremely successful.
In terms of leading us to AGI… GenAI alone won’t reach that. Current ROI is plateauing, and we need to start investing more somewhere else.
It's nothing to do with AI. I didn't believe "I rewrote my application in three weeks!" claims before AI, and I don't believe them now. Most people are not able to evaluate themselves, I don't see why that would have changed.
The goalposts keep getting pushed further and further every month. How many math and coding Olympiads and other benchmarks will LLMs need to dominate before people will actually admit that in some domains it's really quite good.
Sure, if you're a Nobel prize winner or PhD then LLMs aren't as good as you yet, but for 99% of the people in the world, LLMs are better than you at Math, Science, Coding, and every language probably except your native language, and it's probably better at you at that too...
But can they write grammatically correct statements?