These articles are so off from reality to the point that it can't be taken seriously at all. I keep wondering how they keep popping at HN.
At least the author admit he is biased..
At least the author admit he is biased..
This also explains why an independent dev can proclaim a 10x productivity boost or whatever, and actually seem to be showing that - while major companies dumping obscene amounts of $$$ on tokens don't seem to be have much to show for it.
Preposterous. I can tell 5.6 Sol to do something like "optimize this entire subsystem to have no ongoing memory allocations, fix these 5 bugs, implement these 2 features" and go work on other code or do something else entirely, and it does it all flawlessly. All things I could have done myself but nowhere near as fast and right the first time with such thorough test cases. We've reached the point where the AI is usually faster and always less effort overall compared to the old days, even if you have to send a few followup prompts.
Serious question: How do you assess this? It seems to me that it would take very significant time to establish that conclusion.
I find it hard to believe they would spend the effort and money making an AI do tasks like that, but are ALSO willing to spend the effort and time to properly validate the changes. I’m sure many do… but probably not a large percentage.
I mainly look at:
- The overall architecture to make sure it's in line with what I want
- The points where it interacts with certain other systems I'm concerned about
And I test the feature myself in addition to the test cases. Sol has proven to be outstanding at good test cases and having a coherent view on overall architecture integration. I basically never think to myself "this entire implementation is slop, I need to rip it out and refactor it" anymore.
I suspect the model's capacity to translate requirements into contextually appropriate test cases, and insistence on thoroughly covering functionality in tests, is a large part of its strength.
Did you feel the same way about 5.5? The same general sentiment keeps being expressed with every model release. The prior generation is immediately cast aside with a vague “well yes, actually we didn’t mean it last time” attitude.
I think I even heard Theo the T3 guy suggesting that his work comes to halt if he can’t use the latest generation model, despite heaping significant praise on each generation previous and suggesting it does all of his programming tasks.
The boundless energy with which each model generation is said to be revolutionary and leaving everyone behind is tiresome even when incremental progress is real and measurable.
5.5 on Extra High I found to be a huge leap in what it could tackle and how effectively but only up to a point. Still had common failure modes here and there but a big improvement
When I tried 5.6 I wasn't expecting more than an incremental upgrade, but it blew me away with how "smart" it seemed, how it completely sidestepped failure modes I'd been used to dealing with in 5.5, its success rate on the first try for wide-sweeping architectural tasks, and how I've basically never seen a bad hallucination make it through to the final output. It's the first time I've felt like if I left it with a task on the highest setting and it had any way to verify its output, and told it to keep going, it could solve almost anything. This could all be incremental upgrades in many different areas of the model+harness, but the cumulative effect feels like crossing that "escape velocity" threshold. The point where there are no more rough edges and it "just works".
I can't speak to others, but my feelings on all of these have been consistent since I tried them in the first place. I've also heard model providers draw people in at each new release with the model at full performance and then degrade it over time hoping users won't notice, which has seemed plausible anecdotally but I have no hard proof.
But truly I'm mad at the vscode harness. I used to be able to follow the 'thought' of the AI model and steer them when I saw a mistake, it's now way harder to do as multiple research tasks are done in parallel.
They may also be seen as a revolutionary game-changer by people who actually kinda suck at their job.
And a lot of software engineers actually kinda suck at their jobs.
For my own projects, little productivity apps for my own needs, I'm easily 20x more productive. I'm building so many more of them, and at a much higher level of polish. But that's simply because I don't care about the code, long-term maintenance, or anything else. I just need them to do a thing, and if they do the thing, I'm good. I used to make an effort calculation before building these things, but now I just tell Claude "hey, build this", and two hours later it's built well enough to be useful.
OTOH, at work, I'm probably about as fast as I was before, maybe even a bit slower. But the quality of my work has increased quite a bit, because I no longer take the shortcuts I used to take to push things out. I'm spending much more time in planning and in code reviews rather than actually typing code.
Either way, the idea that these obvious changes are just people fooling themselves is, at this point, no longer a reasonable position. And the claim that "AI will run out of money and just be turned off due to the huge operational cost" is so implausible that I would feel ashamed of myself if I used this as a straw man for what AI skeptics believe.
My non-work projects are exercises for the brain and for learning, and if instead I could just automagic them into existence they'd feel hollow and almost cheating.
Sure, let me give you an example. I bought an Xteink X4 recently so I always have something to read with me. I have a huge dump of epubs from Humble Bundle and other sources. I used to just copy them to my e-book readers and then never read anything because it's impossible to find.
I had Claude write a Python script that sorts the books into a file system structure according to Language → Genre → Sub-Genre → Author. Now the device is actually useful, because I can easily find whatever book I want to read, and it has resulted in my reading much more.
> Are you gaining productivity in high-impact things, or stuff that might scratch an itch but you don't really need all things considered?
I technically don't need much. You could lock me in a room and give me stale bread and water once a day. So I'm not sure why you're applying that standard to the usefulness of LLMs, and I'm not sure what you consider "high-impact things". I'm doing high-impact things for me, but maybe you disagree.
So let me ask you this instead: did you write these types of comments in the past, when people manually built software? If I manually wrote a Python script to sort my books, would you post a comment questioning whether that changed anything in my life?
A lot of the criticism leveled against using LLMs in software development is just criticism against software development in general. Do we really need all this software? I guess not. Can we objectively measure the productivity of software developers? Not really.
Which ones? That there is no evidence (not personal anecdotes but the numbers) of productivity increase? Or that our subscriptions are heavy subsidized (then try to compare produced value against real price of tokens)? Or that AI companies are burning tonn of cash without clear plan for monetization?
Article has good points which are worth discussing rather than discarding. We need to bring arguments from both sides of the fense.
The ones that shout about that one METR study, use it to imply that millions of developers are deceiving themselves in believing that tools they use every day are useful, then admit that they don't like and don't use the tools themselves.
The blog tries to cite a 2025 METR study as evidence of no productivity increase, but they ignored the newer 2026 study by the same group that did show a productivity increase with the newer tools.
I also think it's funny that there have become these hard demands for studies and proof of increased productivity. We never saw the same standard of evidence applied to previous advancements like different programming languages or using git for version control. We don't argue with people when they say they're personally more productive in emacs than in VS Code or vice versa. It's only when the topic of AI comes up that the bar for evidence gets raised to the sky
Developers are pushing companies to give them access to AI tools and increase their token budgets.
This opinion piece is more what he wants to happen rather than what's actually happening on the ground
> This is why I'm not using AI. I'm not using AI on moral, ethical grounds.
The funny thing is - they are using AI and they will be using even more AI as it soaks into world economy even more; they just won‘t be using it directly. Well, not for a while at least, until practicality overcomes subjective morals.
So you disagree with the quote from Feynman? (the quote that forms the basis of the article)
I'm genuinely curious to know what part of the article you feel is "so far off from reality"?
> AI will just be turned off.
it's a combination of wishlful thinking and straightup delusion to think that AI will just poof and disappear (maybe the specific LLM approach eventually, but not the general idea of AI/the industry/product as a whole)
So do you disagree that AI companies are in very large amounts of debt? And do you also disagree that debt will have real consequences (and one of those consequences might be either the lights going out or a buyout or merger)?
But if you have a ChatGPT plan and OpenAI is bought by another company (because they are insolvent) ... do you honestly think you will be able to continue using your service without a significant price increase?
I'm very confident that AI will continue to be available no matter what happens, at reasonable prices. We don't have to guess - we know what the open source models take to run, the hardware, the electricity, everything. They are served at a profit today - no-one is subsidizing them. Anyone can run GLM5.2 or (soon) K3 etc - you have capex and opex, you can serve X tokens per second, you sell them at Y price, it's just maths.
AI is a technology, not a company or group of companies.
Dot com bubble was not about the internet. Internet only facilitated it.
The same thing as you, man. Die in the chaos.
Zitron has been predicting an imminent bubble pop since 2023 (although in less and less falsifiable terms), as OpenAI and Anthropic revenues have steadily grown 3-10x per year, but people keep listening. Zitron has done a lot to misinform the public but at some point you have to conclude it's demand-driven, and people won't re-evaluate his credibility because he tells them what they want to hear.
Useful: https://www.theargumentmag.com/p/ais-biggest-critic-has-lost...
And you didn't even mention the debt these companies have accrued in the last 3 years.
References:
https://www.tomshardware.com/tech-industry/big-tech/ai-tech-...
https://finance.yahoo.com/technology/article/techs-ai-debt-b...
https://www.forbes.com/sites/robertszczerba/2026/07/17/bond-...
The question is whether this is a bubble that's going away - whether "AI will just be turned off". A revenue trajectory indicating a vast increase in actual consumer/enterprise usage every year indicates otherwise. This is true regardless of whether or not forward-looking investment becomes too much at some point. If CSPs or AI labs have bought more compute than they can sell at current prices, maybe they'll go bust, but at that point supply and demand mean compute and therefore AI inference becomes much cheaper, which is unlikely to make it go away.
Lol! Feynman's quote is in fact for everyone. Refusing to consider that it might apply to yourself is proof of that!
But imagine that Ctrl+F was just introduced as a feature and you had a bunch of people saying "All these people saying Ctrl+F is increasing their productivity are fooling themselves", it's just a silly thing to say.
Try pointing Fable 5 to almost any CS paper - it can reproduce results in mere hours using maybe couple hundred bucks in retail token prices. This has been obvious to anyone caring to spent even just 30 mins for months now and the writing has been on the wall for more than a year.
2. Does this warrant the unsubsidized cost of AI?
3. Does this justify the harmful externalities of AI?
I don't need personal anecdotes. I want evidence that organisations make more money, get more efficient (and so on) due to AI. Please cite a paper.
You're citing a paper that does not reproduce in the year 2026 and a guy who built a following on doom&gloom projections with zero actual verifiable sources and demanding proof. SMH
The 2026 paper admitted it's different and has issues.
> A lot of software projects don't make money directly. Our company makes money.
That's an evasive answer and you know it. It's about AI making an impact on the bottom line, justifying it's cost.
> A lot of software projects don't make money directly. Our company makes money. This would have cost thousands of dollars paid to a contractor previously with unclear results so yes it's 100% justified.
Thousands of dollars is pocket change even for a small company. That's not the kind of impact AI is supposed to have. You see why I'm not convinced? You also don't talk about the unsubsidised cost of AI nor the negative externalities.
And so if someone like the author is trying to argue the contrary, that leads me to believe they either: (1) have not bothered to spend even 5 minutes validating their belief or (2) do not care to test their beliefs against reality, for whatever reason. Either one is a bit of a red flag.
Just saying.
I'm somewhat agnostic about the whole thing, and I find the utter confidence you display strange and arrogant. Some really good devs say AI is not that great. Some other really good devs say it is great. Why is one side of them automatically delusional? Could they know something you don't?
I wrote and deleted a couple of answers to this - they didn't satisfy, so I'll just be honest.
No, they don't know anything I don't know. There's now only one reason for denying that AI is helpful in software engineering, and that is emotional, be it fear of change or cultural point-scoring.
The only honest answer to the question, "Is modern frontier AI helpful in most software engineering?" is "Yes, of course it is." Anyone answering differently than that is, by now, simply wrong. Why they are wrong may vary, but they are wrong.
Is Fable helpful in that it can write some code that does what I told it to do? Yes, obviously. I don't think anyone is arguing that, because it's trivial to see that if you tell a frontier model to do a coding task, it usually does it well enough.
Is Fable helpful in that over the span of six months of using it, the codebase and my brain's relation to it is in a better state than if I had not used it? Well, I don't know. In this case I understand the code less, and the code is probably in a poor state, but it is further along in the feature list. Whether that tradeoff is a net benefit to whatever metric you care about is debatable.
I think most will agree that the extreme case in which you understand zero and the codebase is spaghetti (true vibecoding), but it has a bunch of features, is not good for most professional contexts, because we have seen this results in a high bugrate and eventual loss of trust in the tool and inability for the AI to maintain invariants due to context size.
So somewhere on the spectrum between no-AI and vibecoding is probably ideal. Where on the spectrum is the argument, and everyone is going to have their own opinions based on their own domains. I don't see this as anyone being wrong or emotional.
So you do not disagree? You agree 100%? Yes, it is trivial to see. Hence my near-anaphylactic reaction to articles like this, which specifically claim that it is NOT useful. Do you just mentally filter that out or something?
> Is [..] the codebase and my brain's relation to it is in a better state than if I had not used it?
But now this is a different question. It will not be the same, that's for sure. With new tools we must learn new patterns, new best practises. I have come around to the opinion that we must collectively move to a higher level of abstraction and accept that we will understand the code less - but I don't agree it must necessarily be worse. That will all depend on the new techniques we discover and apply.
> So somewhere on the spectrum between no-AI and vibecoding is probably ideal. Where on the spectrum is the argument, and everyone is going to have their own opinions based on their own domains. I don't see this as anyone being wrong or emotional.
I think that anyone planting a big flag at 0 (no AI) in 2026, with what we have available, is indeed doing it for emotional reasons, and when they claim there's no benefit being anywhere else, they're wrong. Which is what the article we're talking about is doing - hence my, and other's, scathing reaction.
I just think there is a possibility that being on the AI spectrum is not necessarily better than being at 0. Maybe retaining maximal understanding is better in the long run, even better than minimal, supervised AI usage. Productivity in the long run might be better without it, we just don't know.
It's just not plausible to me that we don't find some way to leverage all that power without fatally undermining everything we're otherwise trying to achieve long-term. It's not plausible to me that skilled, knowledgeable developers like you would ever conclude that AI was never useful, always harmful, and that it is critical to the project's long-term health that you must do this 2,000-file internal refactor entirely by hand.
Just the other day there was data corruption discovered and a relatively urgent fix required. Just a few years ago I would have been working deep into the night to make sure everything was fixed for the morning. Within an hour Fable had it all wrapped up neatly and the fix deployed…
>data corruption...urgent fix required
Do you people hear yourselves?
Ok I guess we just suck, whatever. We and our customers are very happy.
- a lot of folks use ai poorly and innudate you with slop, reducing your own efficiency
- because you can just prompt it doesn't mean you should. many let the machine crank out tokens until it works, rather than thinking about the problem first.
of course, if AI solves my second point automatically, it'll solve the first as well, and we'll call that AGI. But we don't have that yet.
Yeah this article is safe to ignore
This "Anti AI Social Club" is just the latest example, and reality has nothing to do with it. Denying reality might just be the whole point.
It's a very human thing really.