The effects of the technology would have to rival that of.. idk, soap and the locomotive combined? In order for us not to be in a bubble.
It has swallowed nearly all discourse about technology and almost all development, nearly every single area of technology is showing markers of recession.. except AI, which has massively inflated salaries and valuations.
I can’t even think of a scenario where we don’t look back on this as a bubble. What does the tech need to be able to do in order to cover its investment?
Ah, this is such a silly non-issue at this point in time. As long as I have to wait ~10 minutes to get my agent to return code, we are ~10 minutes of compute per request away from a bubble. 10 minutes covers a lot of compute. Nothing we are building right this second will not instantly be eaten up by improving already existing use cases (code, image generation, video generation) for a very long time.
I don't know exactly where AI is going to go, but the fact that I keep seeing this one study of programmer productivity, with like 16 people with limited experience with cursor,uncritically quoted over and over assures me that the anti-AI fervor is at least as big of a bubble as the AI bubble itself.
Rates need to be higher, we are clearly in stagflation outside of AI. Once AI hype cools the stock market will correct. But we need to suck out all of the excess cash, cool consumer demand, which in turn will force businesses to lower their prices to attract customers.
Yeah an exec asked me about the METR study, somewhat alarmed. I told him they were measuring productivity of open source developers. He didn't understand the implications. He was relieved when I informed him that he should still expect productivity gains because his devs were bad enough at their jobs to see a boost.
This is the trouble with AI as an "assistant". If it's good enough to do the job unattended, then there's a big increase in productivity. If it needs constant handholding, maybe not.
It's a good thought exercise, but the opening hinges on a premise that feels obviously wrong: that experts in doing work by hand will do that work faster with AI.
My experience is that, in many cases, people who are very good at doing something by hand are excellent at the process of doing that thing by hand, not generally excellent at doing that thing or talking about that thing or teaching others how to do that thing. And I've found it to be true (sometimes particularly true) for people who have done that thing for a long time, even if they're actively interested in finding new and better ways to do the work. Their skills are fine-tuned for a manual way of working.
Working with AI feels very, very different from writing software by hand. When I let myself lean into AI's strengths, it allows me to move much faster and often without sacrificing quality. But it takes constant effort to avoid the trap of my well-established comfort with doing things by hand.
The people who are best positioned to build incredible things with AI do not have or do not fall into that comfortable habit of manual work. They often aren't expert engineers (yet) in the traditional sense, but in a new sense that's being worked out in realtime by all of us. Gaps in technical understanding are still there, but they're being filled extremely fast by some of the best technology for learning our species has ever created.
It's hard for me to watch them and not see a rapidly approaching future where all this AI skepticism looks like self-soothing delusion from people who will struggle to adjust to the new techniques (and pace) of doing work with these tools.
I think what many people are missing is that everybody values AI companies as if they're special, when I'm seeing customers approach it as a commodity - once a task is done it's done. It has about as much lock-in as a brand of toilet paper, and the economics more like steel production than VPS' or social media.
There are some qualitative differences, but none that seems to last for more than 6 months. On the flip side, the costs are energy & location. Scaling won't bring that down.
In a twist of irony, big tech might structurally reduce their high profit margins because of AI.
I'm bullish on AI, and competition is great for consumers in the end.
But first the bubble has to pop, and it is going to hurt.
This article presents the apparently widespread, but incorrect and, frankly, boring view that "coding" is the bottleneck in development. The following statement summarizes this view best:
> Many knowledge-work tasks are harder to automate than coding, which benefits from huge amounts of training data and clear definitions of success.
which implies that "coding" is not knowledge work. If "coding" is understood as the mere typing of requirements into executable code, then that simply is not a bottleneck and the gains to be had there are marginal (check out Amdahl's law). And if "coding" is to be understood as the more general endeavour of software development, then the people making these statements have no fucking idea what they're talking about and cannot be taken seriously.
I believe it’s a bubble, and that it will burst, but I’m only a little worried about the market.
Yeah, there could be a sell-off of tech, but the people who sold will reinvest elsewhere.
Also, 401ks still exist. Every week people get paychecks and they automatically buy stocks, no matter what the market situation. The only thing that can bring that down is if less people have jobs with retirement plans. If AI busts there will be some amount of re-hiring somewhere to cover it.
Could be scary for people with tech stocks, but less scary with index funds going long term.
18 comments
[ 121 ms ] story [ 94.4 ms ] threador, even worse perhaps the productivity gains will be high.
The effects of the technology would have to rival that of.. idk, soap and the locomotive combined? In order for us not to be in a bubble.
It has swallowed nearly all discourse about technology and almost all development, nearly every single area of technology is showing markers of recession.. except AI, which has massively inflated salaries and valuations.
I can’t even think of a scenario where we don’t look back on this as a bubble. What does the tech need to be able to do in order to cover its investment?
AI is such a thing.
My experience is that, in many cases, people who are very good at doing something by hand are excellent at the process of doing that thing by hand, not generally excellent at doing that thing or talking about that thing or teaching others how to do that thing. And I've found it to be true (sometimes particularly true) for people who have done that thing for a long time, even if they're actively interested in finding new and better ways to do the work. Their skills are fine-tuned for a manual way of working.
Working with AI feels very, very different from writing software by hand. When I let myself lean into AI's strengths, it allows me to move much faster and often without sacrificing quality. But it takes constant effort to avoid the trap of my well-established comfort with doing things by hand.
The people who are best positioned to build incredible things with AI do not have or do not fall into that comfortable habit of manual work. They often aren't expert engineers (yet) in the traditional sense, but in a new sense that's being worked out in realtime by all of us. Gaps in technical understanding are still there, but they're being filled extremely fast by some of the best technology for learning our species has ever created.
It's hard for me to watch them and not see a rapidly approaching future where all this AI skepticism looks like self-soothing delusion from people who will struggle to adjust to the new techniques (and pace) of doing work with these tools.
What do you consider the strengths you're leaning into?
I think what many people are missing is that everybody values AI companies as if they're special, when I'm seeing customers approach it as a commodity - once a task is done it's done. It has about as much lock-in as a brand of toilet paper, and the economics more like steel production than VPS' or social media.
There are some qualitative differences, but none that seems to last for more than 6 months. On the flip side, the costs are energy & location. Scaling won't bring that down.
In a twist of irony, big tech might structurally reduce their high profit margins because of AI.
I'm bullish on AI, and competition is great for consumers in the end.
But first the bubble has to pop, and it is going to hurt.
(@dang, why can't I link it under https://news.ycombinator.com/item?id=45161656 ?)
> Many knowledge-work tasks are harder to automate than coding, which benefits from huge amounts of training data and clear definitions of success.
which implies that "coding" is not knowledge work. If "coding" is understood as the mere typing of requirements into executable code, then that simply is not a bottleneck and the gains to be had there are marginal (check out Amdahl's law). And if "coding" is to be understood as the more general endeavour of software development, then the people making these statements have no fucking idea what they're talking about and cannot be taken seriously.
Yeah, there could be a sell-off of tech, but the people who sold will reinvest elsewhere.
Also, 401ks still exist. Every week people get paychecks and they automatically buy stocks, no matter what the market situation. The only thing that can bring that down is if less people have jobs with retirement plans. If AI busts there will be some amount of re-hiring somewhere to cover it.
Could be scary for people with tech stocks, but less scary with index funds going long term.
https://metr.org/blog/2025-07-10-early-2025-ai-experienced-o...
This feels like confusing the economy with the markets. Now, a market crash would cause economic fallout, but the markets are not the economy.