Yeah but the investments arent aiming for churning out SaaS apps. Its to automate large swathes of intellectual labour. Of which only SWE has been cracked yet. There is a question mark as to if the others will crack. If they arent then these investments will collapse from speculation down to reality. That possibility is what is being discussed here
As to whether that will happen, I think that risk is real. Because claude code isnt made by the generalozed capabilities of the tech but by good old non-generalozable hueristics and rule based engines. I dont think that will scale to other feilds at the factor these investments assume. Its the bitter lesson again. It scales with deliberate and specific design, not data, so it wont scale
We learnt this with ibm watson. Deepblue achieved chess supremacy but the last mile wasnt data driven, it was heiristic driven, and so watson, its successor, couldnt scale/generalize.
My prediction is that this speculation on LLMs with harnesses will collapse since they wont scale. We'll have another winter where the reasearchers will be leaft alone long wnough to come up with the next breakthrough (probably game theory based data driven agency) which might then create what this hypecycle is speculating
There’s argument to be made that SWE hasn’t been cracked either. The latest models are great at coding medium sized applications, but figuring out the requirements and consolidation of domain knowledge is something still lacking
We will find out how much of work is given to people just so that there's a person/company associated with a technical decision. I personally think this might be quite high.
Exactly. I build automation tools for my company which have improved productivity quite a lot and put precisely zero people out of a job. Partly we find other things for them to spend time on, and partly it just turns out that we like to have humans doing jobs.
It is cliche at this point that HN is the place you go to hear software developers reduce all of the world's problems into simple algorithmic arguments which for some reason never actually solve anything. Not shocked that we are similarly incapable of understanding that algorithmically replacing a software developer isn't easy just because we think we know what the job is.
you mentioned a very good point about scalability. we're seeing alot of productivity gains, but only from SWEs, which are but a very small segment of the global economy. all other economic use cases require thorough last-mile development and iteration that is not too different with current automation tools.
A friend who is a psychologist was telling me he thinks in another year or two insurance companies will insist people see an AI therapist first before being willing to pay for a real person.
Tracks. This is that same speculation that AI will be good enough. Wonder what a crash in that isecase will look like? Increased suicide rates? More instances of psychosis? It might not wven be directly measurable or easily traceable to AI therapists. Would suck
LOL, the same AI that has landed companies in court defending themselves against wrongful death lawsuits for helping someone convince themselves suicide is the right answer, and even encouraging them? That AI? I am unclear that any insurance company is going to want a piece of that action anytime soon.
What you've just told me is that psychologists, just like SWEs, are prone to thinking they know how business works but in fact know fuck all.
All those automaton tools will eventually be initially one-shotted and then monitored by LLMs though. There probably won't be a "last mile" per se; just constant tweaking and optimizations throughout, within a feedback loop.
What your describing is iterating in the last mile. Your assuming that the AI will iterate in the last mile with the same efficacy that it iterates before that. I think that will fail. Thats the bitter lesson, that adhoc solutions to the last mile (rather than generalozed solutions that scale with training data) asymtotally stall and so dont scale.
Meaning claude code wont be able to make a "claude video editing" or "claude accounting" with the current tech. Human experts will need to encode their knowledge into it for the last mile and that wont scale the way these speculations expect
We aren't seeing productivity gains in software either. What we are seeing is a lot of people who claim to be more productive, but in fact are building piles of tech debt that will fall over before long. But hey, they're building that tech debt really fast!
> but in fact are building piles of tech debt that will fall over before long
This is speculation as well. Its well founded but speculation nonetheless. Youre speculating things will stay the way they have till now.
I do see your point but what makes me consoder the other side is that ive been building an app that reaches ~10k LOC, purely with opus, no code review at all, and it hasnt hit any tech debt issues that i havent easily been able to address. Setting up good context management meant that claude could just figure things out itself.
And for reference this is an app that manages an ethernet camera, runs vehicle detection on the stream, and surfaces the detections on an ipad for operators to inspect and annotate for cellphone usage, so not trivial. Needed good architechtong and design from my end, but it was honestly scarily easy. So idk what the threshold for tech debt crash is but it wasnt there
While I don't agree with your premise at all, even if it could one shot a SaaS product (a statement so vague it's meaningless) I don't think there's much of an argument for why that's economically useful. A lot of SaaS has free software/open source equivalents anyway (how else do you think the clanker's are able to plagiarize it?). People still pay for Office even though you could easily use LibreOffice, or GitHub when you could self host Forgejo. It's like when anthropic made a big deal out of making a broken compiler. Neat, so, after ingesting all of open source and burning a trillion tokens you ended up with something worse than what's already out there; and instead of doing something economically useful like giving a person money to build it or supporting the open source ecosystem, you're just wasting energy on datacenters.
So good at it that I’m right now in the process of building instead of buying.
Here’s how that plays out in the economy:
- My company spent $50 on my tokens to build this internal tool
- Anthropic spent $XXX to deliver those tokens to me.
- The company I was going to buy the tool from lost $XX,XXX per year that I would have paid them.
I dunno, kind of sounds like the economy just got smaller.
I could usually accept the idea that software getting cheaper generally increases demand for software and expands the economy surrounding it, but I’m not sure if we have precedent for what happens when software becomes positively worthless.
I certainly can’t give you a better answer for my company than “it depends” or “I don’t really know.”
The company could just be happy to have better margins and be happy the stock finally went up. It might literally do nothing with them or do something economically unproductive like buy back stock.
What I can tell you with certainty is that we aren’t going to hire anyone else or launch any other product as a result. Our business just isn’t at that level of growth potential.
Perhaps we can surmise that money going to shareholders can grow the economy. They’ve got more money to reinvest in other stuff.
But then again, if everyone can shart out a SaaS app with $50 in tokens, what software companies will they want to invest in?
AI gives me that feeling of “what happens to bakers and butchers when the supermarket gets invented and they decide to sell bread and meat at or below cost?”
Every company has a list of >WACC IRR projects that it can spend saved money on. If not, it’s a cash cow company that wasn’t growing in the first place and will allow shareholders to use the saved cash for other economically expanding projects.
What companies can expand if the income of consumers is shrinking. This is the scary bit to me — AI crashes and takes the economy with it, or; AI succeeds as promised and people go unemployed and crash the economy.
The only path that isn’t disastrous is threading the needle of “just right” productivity gains. The people in charge aren’t smart enough to give me warm fuzzy feelings on that.
Economy != software companies. Maybe there needs to be a capital rotation out of tech? That’s speaking beyond my expertise though (and imo is a little too doomer). Continuing the hypothetical through: Healthcare, industrials, financial services and many other verticals have plenty of growth opportunities.
Otherwise it would probably be the software companies that are the most focused on last-mile details (where AI in my experience has the most trouble). I expect that as consumers are faced with more and more AI slop SaaS they will be increasingly willing and able to pay for quality.
And if the company didn't need $XX,XXX*0.90 (or more) that you would have paid them to further develop their product and stay in business? If that other company now paid their own $50 in tokens? Maybe the overall flow of money in the economy went from $XX,XXX (you) + $XX,XXX (them) + $(not much, AI didn't exist yet) equals or is greater to $50 + $50 + ($xbillions in AI)? Dunno.
wouldn't take long for them to find out with the developer bragging about it. he one-shot vibe-coded a sass app (keyword: one-shot), that says alot about the quality.
> Meanwhile AI has gotten so good it can just about one shot a SaaS app.
There isn't a direct correlation between AI improvement or stagnation and whether or not the amount being spent by AI labs and the associated ecosystem will result in a financial crash.
Look into the history of railroads and the internet itself to see how massive levels of investment can result in economic crashes even when the thing being invested in produces real, widespread societal value.
One could argue that one of the nightmare economic scenarios for AI is actually that it gets too good too fast and results in a wipeout of the white collar worker that we are currently nowhere near ready to deal with given how propped up our economy is on consumer spending.
difference this time is they have "fiat money" and money printer. Market and all inv. bankers knows that in major crash they will print unlimited amounts so back to same prices or near them. printer is still printing and it's only goes to selected investments
As if they did not back then. Fiat is just simpler to work with, but one can pull a bubble without it just fine. Anyone forgot the railroad crash of 1873? The tulip mania of 17th century?
As to whether that will happen, I think that risk is real. Because claude code isnt made by the generalozed capabilities of the tech but by good old non-generalozable hueristics and rule based engines. I dont think that will scale to other feilds at the factor these investments assume. Its the bitter lesson again. It scales with deliberate and specific design, not data, so it wont scale
We learnt this with ibm watson. Deepblue achieved chess supremacy but the last mile wasnt data driven, it was heiristic driven, and so watson, its successor, couldnt scale/generalize.
My prediction is that this speculation on LLMs with harnesses will collapse since they wont scale. We'll have another winter where the reasearchers will be leaft alone long wnough to come up with the next breakthrough (probably game theory based data driven agency) which might then create what this hypecycle is speculating