AI Pullback Has Officially Started(planetearthandbeyond.co)
planetearthandbeyond.co
AI Pullback Has Officially Started
https://www.planetearthandbeyond.co/p/ai-pullback-has-officially-started
65 comments
> But really, typing up the code has always been the low-skill part of the job!
I once heard this put in the context of engineer-code vs software-developer-code:
A professional software developer's skill isn't writing working code (anyone with enough time and intelligence can do that), but rather writing maintainable, efficient working code.
I once heard this put in the context of engineer-code vs software-developer-code:
A professional software developer's skill isn't writing working code (anyone with enough time and intelligence can do that), but rather writing maintainable, efficient working code.
...except a significant number of us programmers also think it's BS. The pullback also includes less use in development automation lately as well.
I'm actually with you on that one to be honest.
There's a few bits of information from the original sources that's left out:
- The METR paper surveyed just 16 developers to arrive at their conclusion. Not sure how that got past review. [0]
- The finding from the MIT report can also be viewed from a glass 5% full perspective:
> Just 5% of integrated AI pilots are extracting millions in value. > Winning startups build systems that learn from feedback (66% of executives want this), retain context (63% demand this), and customize deeply to specific workflows. They start at workflow edges with significant customization, then scale into core processes. [1]
[0] https://arxiv.org/abs/2507.09089
[1] https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Bus...
- The METR paper surveyed just 16 developers to arrive at their conclusion. Not sure how that got past review. [0]
- The finding from the MIT report can also be viewed from a glass 5% full perspective:
> Just 5% of integrated AI pilots are extracting millions in value. > Winning startups build systems that learn from feedback (66% of executives want this), retain context (63% demand this), and customize deeply to specific workflows. They start at workflow edges with significant customization, then scale into core processes. [1]
[0] https://arxiv.org/abs/2507.09089
[1] https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Bus...
AI creates the most spectacular happy path demos. It’s hard not to extrapolate to infinity when you see it.
People have a bias to want to believe something works in all cases, when it seemingly offers benefits to them. Especially when there’s a sunk investment involved.
This was always kind of a problem with the “this will make icky programmers obsolete” techs. Like, so did MS Access and a couple generations of click-and-drag ‘no-code’ stuff. Not to mention Rails; remember when everyone thought that would radically increase productivity? I’m pretty sure that was entirely because it was well-suited to “make a todo list/fake twitter/whatever in half an hour” demos.
> To stem the backlash, many journals and universities are starting to resist or have stopped using AI altogether in the peer review process.
… Excuse me, they were doing _what_? The world has gone mad. Unless you think a chatbot that doesn’t know how many ‘r’s are in strawberry is your peer you shouldn’t be using it for peer review, bloody hell.
At least in tech there’s still usually human code review which catches the worst of the magic robot-generated nonsense.
… Excuse me, they were doing _what_? The world has gone mad. Unless you think a chatbot that doesn’t know how many ‘r’s are in strawberry is your peer you shouldn’t be using it for peer review, bloody hell.
At least in tech there’s still usually human code review which catches the worst of the magic robot-generated nonsense.
I find these sorts of takes to be tiresome.
It is absolutely true there is a lot of hype around AI. Also it is true that many AI companies try to shove AI into everything without necessarily thinking wherer it is a good idea or whether it useful (talking to you Google).
Notwithstanding this it is absolutely clear how transformational the technology is. For low skill tasks it can certainly substitute people and save a lot of time. For harder things one has to be more careful and the right model have to be used, i.e. it is not a silver bullet but just a powerful tool which means it needs to be used in conjunction with other tools
You’re gonna keep seeing takes like this until expectations meet the reality of the tools, because the hype is absolutely insane right now. I think it’s good to push back against the current narrative.
For every employee using LLM for productivity you have 50 who use it to bullshit their way up the ladder, generate overly verbose emails, reports, bug reports, &c.
My wife's team spent 20+ man hours analyzing and trying to fufil the requests of one of their biggest customer, in the end it turned out to be a fully llm generated feature request email from someone who didn't quite understand the product in the first place...
When you save one hour on a coding task somewhere someone spends two hours trying to parse some bullshit email or report. I'm convinced it's a net negative overall
My wife's team spent 20+ man hours analyzing and trying to fufil the requests of one of their biggest customer, in the end it turned out to be a fully llm generated feature request email from someone who didn't quite understand the product in the first place...
When you save one hour on a coding task somewhere someone spends two hours trying to parse some bullshit email or report. I'm convinced it's a net negative overall
Most business tasks are not low skilled language tasks that are so easy to automate. I am helping rebuild a business right now that has a call center and LLMs are basically useless for us.
It is actually kind of shocking how I can go home and learn about quantum computing from LLMs but find them useless for simple business processes. This though I think exemplifies the entire mistake of the bubble. Most business processes don't benefit at all from understanding the Hamiltonian of a system. Most business processes are simple tasks that are the end result of previous automation. In practice, most business processes in 2025 are simple processes done by a human who can deal with the random uncertainty and distribution shift that inevitably comes up. Exactly what a language isn't good but it is even beyond that. So much of what a customer service agent for example is doing is dealing with uncertainty and externality. There is the process that LLMs aren't good anyway but then there is the human judgement on when to disregard the process because of various externalities. The trivial business processes LLMs would be good at automating have already been automated years ago or the business went out of business years ago. AGI would in theory be amazing at all this too but we don't have AGI. We have language models that have a very limited use case beyond an interactive version of Wikipedia. I love the interactive version of Wikipedia but it is not worth trillions of dollars.
It is actually kind of shocking how I can go home and learn about quantum computing from LLMs but find them useless for simple business processes. This though I think exemplifies the entire mistake of the bubble. Most business processes don't benefit at all from understanding the Hamiltonian of a system. Most business processes are simple tasks that are the end result of previous automation. In practice, most business processes in 2025 are simple processes done by a human who can deal with the random uncertainty and distribution shift that inevitably comes up. Exactly what a language isn't good but it is even beyond that. So much of what a customer service agent for example is doing is dealing with uncertainty and externality. There is the process that LLMs aren't good anyway but then there is the human judgement on when to disregard the process because of various externalities. The trivial business processes LLMs would be good at automating have already been automated years ago or the business went out of business years ago. AGI would in theory be amazing at all this too but we don't have AGI. We have language models that have a very limited use case beyond an interactive version of Wikipedia. I love the interactive version of Wikipedia but it is not worth trillions of dollars.
Im curious I've seen very different results for automation of business processes. Could you share an example of what you're dealing with?
takes are there cause AI is so bad that no one reads anything else orher than stories about how AI is so bad :)
> For low skill tasks it can certainly substitute people and save a lot of time.
I mean, per the article, only if you don’t care about correctness. There aren’t actually that many use cases where correctness doesn’t matter at all.
I mean, per the article, only if you don’t care about correctness. There aren’t actually that many use cases where correctness doesn’t matter at all.
Yes. I think the hype is crazy and also the tools are awesome. Like the other day i had all this different documentation in different files. .tex, .md, .docx, .xlsx, it was a mess and I needed to put it together into a summary document that could be shared with stakeholders because they aren't going to read through all that and make sense of it. So I dropped them in a folder and asked claude code to do it and 5 minutes later I had something I could edit a bit and send out. That would have been at least half a days worth of work. Did it fudge some bits? yes but then i just fixed it and everything is fine.
to me i think all the hype comes from promises of C3P0s and R2D2s instead pitching it as building computational tools that make you more efficient or give you new ways to model your ideas inside a computer.
to me i think all the hype comes from promises of C3P0s and R2D2s instead pitching it as building computational tools that make you more efficient or give you new ways to model your ideas inside a computer.
AI (LLM) is useful for coding and I use it to lookup various articles or websites and summarize.
Use it where it works.. ignore the agents hype and other bullshit peddled by 19yo dropouts.
Unlike the 19yo dropouts of the 2010s these guys have brain rot and I don’t trust them after having talked to such people at start up events and getting their black pill takes. They have products that don’t work and lie about numbers.
I’ll trust people like Karpathy and others who are genuinely smart af and not kumon products.
Use it where it works.. ignore the agents hype and other bullshit peddled by 19yo dropouts.
Unlike the 19yo dropouts of the 2010s these guys have brain rot and I don’t trust them after having talked to such people at start up events and getting their black pill takes. They have products that don’t work and lie about numbers.
I’ll trust people like Karpathy and others who are genuinely smart af and not kumon products.
On this website on another thread there is a principal software engineer at Microsoft who wrote an essay on how agent systems are amplifying all of the employees productivity massively even on large complex tasks.
I mean, MS is _deep_ into this; “company who is going to plow literally one hundred billion dollars into thing in the next year says thing is just great” is not _terribly_ convincing, honestly.
Now the question is whether that’s true (and thus should be objectively measurable) or if he is bullshitting because Microsoft invested so much money in it it just has to work.
yes, I believe this. Microsoft is deploying an army of devs to HN to tout AI because they are invested in at cost of billions of dollars per year - HN AI Bubble :)
You don't need a coordinated disinformation campaign for employees to drink the Kool-Aid and evangelize.
all it takes is significant emotional investment for someone to become a bit blind to reality.
all it takes is significant emotional investment for someone to become a bit blind to reality.
“Principal” at Microsoft (or Oracle) is “Senior” or “Staff” everywhere else just fyi
I often wonder how much code people are generating at once to get such poor quality. I'm usually doing one behavior on a class/classes at a time and get great results. It's a bit longer and a little more tedious but by the time I am done the context window is refined enough that it can write my tests with ease.
I'm not saying AI is living up to the "hype" or "expectations" - it would largely depend on how you quantify the hype or expectations. Most rational would be to consider how much money is funneled into vs how much ROI would it have within some time range in the future, e.g. 10 years. A wise investor would look ahead 10 years, balance benefits, potential and risks. By that metric it could be too early to say if it's paying off even if it's objectively clearly bringing 10x more expense than income.
But the metrics or facts without context or deeper explanations also don't mean much in that article.
> 95% of AI pilots didn’t increase a company’s profit or productivity
If 5% do that could very well be enough to justify it, depending on for which reasons and after how much time the pilots are failing. It's widely touted that only 5% of start ups succeed, yet start ups overall have brought immense technological and productivity gains to the World. You could live in a hut and be happy, and argue none of it is needed, but none the less the gains by some metrics are here, despite 95% failing.
The article throws out numbers to make a point that it wanted to make, but fails to account for any nuance.
If there's a promising new tech, it makes sense that there will be many failed attempts to make use of it, and it makes sense a lot of money will be thrown in. If 5% succeed, it takes 1 million to do 1 attempt, but the potential is 1 billion if it succeeds, it's already 50x return.
In my personal experience, if used correctly it increases my own productivity a lot and I've been using AI daily ever since GPT 3.5 release. I would say I use it during most of what I do.
> AI Pullback Has Officially Started
So I'm personally not seeing this at all, based on how much I personally pay for AI, how much I use it, and how I see it iteratively improving, while it's already so useful for me.
We are building and seeing things that weren't realistic or feasible before now.
But the metrics or facts without context or deeper explanations also don't mean much in that article.
> 95% of AI pilots didn’t increase a company’s profit or productivity
If 5% do that could very well be enough to justify it, depending on for which reasons and after how much time the pilots are failing. It's widely touted that only 5% of start ups succeed, yet start ups overall have brought immense technological and productivity gains to the World. You could live in a hut and be happy, and argue none of it is needed, but none the less the gains by some metrics are here, despite 95% failing.
The article throws out numbers to make a point that it wanted to make, but fails to account for any nuance.
If there's a promising new tech, it makes sense that there will be many failed attempts to make use of it, and it makes sense a lot of money will be thrown in. If 5% succeed, it takes 1 million to do 1 attempt, but the potential is 1 billion if it succeeds, it's already 50x return.
In my personal experience, if used correctly it increases my own productivity a lot and I've been using AI daily ever since GPT 3.5 release. I would say I use it during most of what I do.
> AI Pullback Has Officially Started
So I'm personally not seeing this at all, based on how much I personally pay for AI, how much I use it, and how I see it iteratively improving, while it's already so useful for me.
We are building and seeing things that weren't realistic or feasible before now.
What if switching to 3 ply toilet paper (from the dreaded 1 ply) in employee bathrooms increased productivity in 5% of companies. We could also apply the same logic that these 5% could also produce 50x returns.
So firstly to be comparable, you'd have to make a claim that we are in a 3 ply toilet paper overhype cycle that is about to pullback and make a claim that 95% of the companies that use it, fail. Then I would come in and ask about the 5%.
I'm not sure what your point is, but I do absolutely think that getting proper toilet paper offers 50x returns and everyone should prioritize it.
Better yet, get a bidet :)
The point is, the above logic depends on the part about "1 million" and "1 billion", that is, it assumes a thousand times profit for the 5% of companies that succeed in getting productivity improvements from AI or toilet paper. Eh, but I guess TP is cheap, so maybe you're going along with that.
More so I'm showing how it's not evidence that AI pullback has started.
To prove, that AI is a overhyped bubble and won't bring in the ROI expected, you'd need to show that these companies won't bring expected ROI within a longer timeframe.
To prove, that AI is a overhyped bubble and won't bring in the ROI expected, you'd need to show that these companies won't bring expected ROI within a longer timeframe.
Time is ticking and running out. You do realise future numbers have to be discounted heavily for the required rate of return given the risk, right?
This means the free cash flow to the firm OAI generates will have to be huge, given the negative cash outflows to date.
This means the free cash flow to the firm OAI generates will have to be huge, given the negative cash outflows to date.
Then to prove that it's a bubble and a failing bubble, you should bring out the specific numbers if you do have access to them. I'm just saying the numbers outlaid in the blog post are not evidence alone and brought examples where similar numbers could be observed while something is still highly successful.
5% succeeding is abysmal for an industry where a trillion dollars or more is invested in.
And that’s ignoring the rampant copyright infringement, the exploding power use and accompanying increase in climate change, the harm it already does to people who are incapable of dealing with a sycophantic lying machine, the huge amounts of extremely low quality text and code and social media clips it produces. Oh and the further damage it is going to do to civil society because while we already struggled with fake news, this is turning the dial not to 11 but to 100.
And that’s ignoring the rampant copyright infringement, the exploding power use and accompanying increase in climate change, the harm it already does to people who are incapable of dealing with a sycophantic lying machine, the huge amounts of extremely low quality text and code and social media clips it produces. Oh and the further damage it is going to do to civil society because while we already struggled with fake news, this is turning the dial not to 11 but to 100.
Sure, it can do plenty of harm, but it doesn't mean that AI pullback has officially started or that it won't bring in the ROI for the investors.
lolol "I still use it, so clearly everyone else still loves it and thinks it's amazing"
I’ve been using AI coding systems for quite some time, and have worked in neural networks since the 90’s. The advancements are, frankly, almost as crazy as 90’s neural net devotees like me were claiming could be possible in the eventual future.
That said, the non-tech-executive/product-management take on AI has often been an utter failure to recognize key differences between problems and systems. I spend an inordinate amount of time framing questions in terms of promises to customers, completeness, reproducibility, and contextual complexity.
However, for someone in my role, building and ideating in innovation programs, the power of LLM assisted coding is hard to pass up. It may only get things 50% of the way there before collapsing into a spiral of sloppy overwrought code, but we often only need 30-40% fidelity to exercise an idea. Ideation is a great space for vibe coding. However, one enormous risk in these approaches is in overpromising the undeliverable. If folks don’t keep a sharp eye on the nature of the promises they’re making, they may be in for a pretty wild ride; with the last “20%” of the program taking more than 90% of the calendar time due to compression of the first “80%” and complication of the remainder.
We’re going to need to adjust. These tools are here to stay, but they’re far from taking over the whole show.
That said, the non-tech-executive/product-management take on AI has often been an utter failure to recognize key differences between problems and systems. I spend an inordinate amount of time framing questions in terms of promises to customers, completeness, reproducibility, and contextual complexity.
However, for someone in my role, building and ideating in innovation programs, the power of LLM assisted coding is hard to pass up. It may only get things 50% of the way there before collapsing into a spiral of sloppy overwrought code, but we often only need 30-40% fidelity to exercise an idea. Ideation is a great space for vibe coding. However, one enormous risk in these approaches is in overpromising the undeliverable. If folks don’t keep a sharp eye on the nature of the promises they’re making, they may be in for a pretty wild ride; with the last “20%” of the program taking more than 90% of the calendar time due to compression of the first “80%” and complication of the remainder.
We’re going to need to adjust. These tools are here to stay, but they’re far from taking over the whole show.
There are a lot of people invested in AI, so they are cheerleaders. There are way more people who didn't invest, who are sour grapes and want to see it fail. I'm neither of these people, but it's a democracy after all. I think AI is due for another winter.
I think it's not as much a democracy as a market. We sometimes say people vote with their wallet on products, so I see where you come from.
Still, in this case I think a market analogy fits better. There are people who want it and people who don't want it. If the people with a lot of money (to manage for companies) want it, this will move the balance. If it eventually moves it enough remains to be seen. Decisions can be made with too much excitement and based on overpromises, but eventually someone will draw a bottom line under (generative) AI, the one where currently the huge amount of money gets pumped into. Either will generate generate value that people pay for and the investors make a profit or not. Bubbles and misconceptions can extend the time when the line is drawn, but eventually it will be.
If LLM and generative is generally creates value, or not, I cannot say. I am sure that the more specialised AI solutions that are better described as machine learning does create this value in their special use cases and will stay.
Still, in this case I think a market analogy fits better. There are people who want it and people who don't want it. If the people with a lot of money (to manage for companies) want it, this will move the balance. If it eventually moves it enough remains to be seen. Decisions can be made with too much excitement and based on overpromises, but eventually someone will draw a bottom line under (generative) AI, the one where currently the huge amount of money gets pumped into. Either will generate generate value that people pay for and the investors make a profit or not. Bubbles and misconceptions can extend the time when the line is drawn, but eventually it will be.
If LLM and generative is generally creates value, or not, I cannot say. I am sure that the more specialised AI solutions that are better described as machine learning does create this value in their special use cases and will stay.
Painting opposition as "sour grapes" is an extraordinarily bad faith take.
I dont want to be sold ai girlfriends
If that makes me a sour grape, so be it.
If that makes me a sour grape, so be it.
I'm fine with AI girlfriends and boyfriends.
At least those are not directly harmful to me.
I'm slightly less fine when my time is wasted by some generated bullshit.
And not at all fine when some vibed some product and ignored basic good practises on security and so on.
At least those are not directly harmful to me.
I'm slightly less fine when my time is wasted by some generated bullshit.
And not at all fine when some vibed some product and ignored basic good practises on security and so on.
I am, due to broad index funds, more invested in ‘AI’ than I would like. Still think it’s largely snake oil.
We should expect pullbacks, fuckups, plans failing, and rollouts getting canned. It's part of how humans do things. Its actually a pretty effective optimization algorithm.
I'd bet that some sort of exponentiate the learning rate until shit goes haywire then rollback the weights is actually probably a fairly decent algorithm (something like backtracking line search).
I'd bet that some sort of exponentiate the learning rate until shit goes haywire then rollback the weights is actually probably a fairly decent algorithm (something like backtracking line search).
People roll out a complex and powerful technology without understanding the technology fully, what evals are or updating process to account for the tech, and the rollout fails, news at 11.
Seriously though, "AI fucks up" is a known thing (as is humans fuck up!) and the people who are using the tech successfully account for that and build guardrails into their systems. Use version control, build automated tests (e2e/stress, not just unit), update your process so you're not incentivizing dumb shit like employees dumping unchecked AI prs, etc.
Seriously though, "AI fucks up" is a known thing (as is humans fuck up!) and the people who are using the tech successfully account for that and build guardrails into their systems. Use version control, build automated tests (e2e/stress, not just unit), update your process so you're not incentivizing dumb shit like employees dumping unchecked AI prs, etc.
If the tech only worked for coding it would be one thing. But it’s advertised as a cure for anything and everything and so people are using it for that. And you can’t build automated tests for that.
I am a big AI booster but I agree that using agents for tasks unsupervised without either rigorous oversight or strong automated constraints is a mistake.
Imagine comparing a human fuck up to an AI one. Lol.
it's been grinding my gears so much lately that people keep trying to compare "blurry jpeg machines" to human intelligence and development.
llms don't learn. nor do they operate with any sort of intent towards precision.
we can develop around, plan for and predict most common human errors. also, humans typically get smarter and learn from their mistakes.
llms will go on making the same ridiculous mistakes, confidently making up bullshit frameworks methods and code, and no matter how much correction you try to offer, they will never get any better until the next multi-billion dollar model update. and even then, it's more of a crossed finger situation than an inevitability improvement and growth.
I hate hate hate hate hate that AI seems to be increasing Dunning Krueger's effect on all our lives...
llms don't learn. nor do they operate with any sort of intent towards precision.
we can develop around, plan for and predict most common human errors. also, humans typically get smarter and learn from their mistakes.
llms will go on making the same ridiculous mistakes, confidently making up bullshit frameworks methods and code, and no matter how much correction you try to offer, they will never get any better until the next multi-billion dollar model update. and even then, it's more of a crossed finger situation than an inevitability improvement and growth.
I hate hate hate hate hate that AI seems to be increasing Dunning Krueger's effect on all our lives...
I have seen 30 years of human fuckups, it is infinitely worse than AI fuck ups so you are right, cannot be compared, humans are so much worse
there are roughly 2.09% of SWEs that actually know what they are doing so this 97.91% generally prodces garbage (after 30 years doing this shit I have once experiencing being brought it to a project (I have been working as a consultant for a long time now) and went “wow, now this is beautiful codebase!”
there are roughly 2.09% of SWEs that actually know what they are doing so this 97.91% generally prodces garbage (after 30 years doing this shit I have once experiencing being brought it to a project (I have been working as a consultant for a long time now) and went “wow, now this is beautiful codebase!”
You have to look at where the bullet holes aren't on surviving planes to know where to reinforce them...
aka you don't maybe think tha - as an outside consultant - the nature of the job means you'd rarely be brought in to fix "beautiful codebases"...?
aka you don't maybe think tha - as an outside consultant - the nature of the job means you'd rarely be brought in to fix "beautiful codebases"...?
certainly! but you see so much you stay in the industry long enough. and hear other people’s stories. the most common one - “just got a new gig at ____, wow the codebase is a mess.”
I probably worked with 300-400 SWEs directly and of them there is only one I’d trust to write code if my life depended on it. and I think that is likely in-line with how many SWEs are actually great at their jobs
I probably worked with 300-400 SWEs directly and of them there is only one I’d trust to write code if my life depended on it. and I think that is likely in-line with how many SWEs are actually great at their jobs
Articles like this keep popping up because they're catnip to those who hate AI or feel threatened by it. On coding, I routinely see people trashing vibe coding and jumping on the slightest mistake agents may make, never mind that human devs screw up all the time. And write-ups citing stats on AI coding tend to be written by folks who either don't code for a living or never earnestly tried it.
I use Claude Code regularly at work and can tell it is absolutely fantastic and getting better. You obviously need to guide it well (use plan mode first) and point to hand coded stuff to follow, and it will save you enormous amount of time and effort. Please don't put off trying AI coding out after reading misinformed articles like this.
I use Claude Code regularly at work and can tell it is absolutely fantastic and getting better. You obviously need to guide it well (use plan mode first) and point to hand coded stuff to follow, and it will save you enormous amount of time and effort. Please don't put off trying AI coding out after reading misinformed articles like this.
Sorry, but your anecdote isn’t very convincing in comparison to the data shared in the article.
I think devs have a natural inclination to resist a seismic shift in their industry, which is understandable.
However I agree that a lot of this stuff is FUD and AI dev is like a new skill, it takes time to master. It took me a few months but I’m comfortably more productive and having a more fun time at work with Claude Code
However I agree that a lot of this stuff is FUD and AI dev is like a new skill, it takes time to master. It took me a few months but I’m comfortably more productive and having a more fun time at work with Claude Code
To people who aren't programmers, there isn't really the same kind of easily-verified use case. Most people can't tell at a glance that a business proposal or email is full of errors they need to correct, thus the stuff causes even more damage.
Unfortunately, programmers, as a rule, aren't terribly good at listening to the experiences and perspectives of non-coders, so I don't see this dynamic changing anytime soon.