The most credible argument against AI is really the expense involved in querying frontier models. If you want to strengthen the case for AI-assisted coding, try to come up with ways of doing that effectively with a cheap "mini"-class model, or even something that runs locally. "You can spend $20k in tokens and have AI write a full C compiler in a week!" is not a very sensible argument for anything.

How much would it cost to pay developer to do this??

It’s hard to say. The compiler is in a state that isn’t useful for anything at all and it’s 100k lines of code for something that could probably be 10k-20k.

But even assuming it was somehow a useful piece of software that you’d want to pay for, the creator setup a test harness to use gcc as an oracle. So it has an oracle for every possible input and output. Plus there are thousands of C compilers in its training set.

If you are in a position where you are trying to reverse engineer an exact copy of something that already exists (maybe in another language) and you can’t just fork that thing then maybe a better version of this process could be useful. But that’s a very narrow use case.

The cost argument is a fallacy, because right now, either you have a trained human in the loop, or the model inevitably creates a mess.

But regardless, services are extremely cheap right now, to the point where every single company involved in generative AI are losing billions. Let’s see what happens when prices go up 10x.

zero

because they tell you to stop being so stupid and run apt install gcc

Because hardware costs never goes down and energy efficiency never go up overtime?

Whatever the value/$ is now, do you really think it is going to be constant?

If hardware industry news is any indication, hardware costs aren't going to be going down for GPUs, RAM, or much of anything over the next 3-5 years.

Maybe, but I seriously doubt that new DRAM and chip FABs aren't being planned and built right now to push supply and demand to more of an equilibrium. NVIDIA and Samsung and whoever else would love to expand their market than to wait for a competitor to expand it for them.

How long do you think it takes for those factories to go from nothing to making state-of-the-art chips at a scale that's large enough to influence the supply even by 1%?

There are plenty of them being built, yes. Some of them will even start outputting products soon enough. None of them are gonna start outputting products at a scale large enough to matter any time soon. Certainly not before 2030, and a lot of things can change until then which might make the companies abandon their efforts all together or downscale their investments to the point where that due date gets pushed back much further.

That's not even discussing how easier it is for an already-established player to scale up their supply versus a brand-new competitor to go from zero to one.

If you keep digging, you will also find that there's a small but vocal sock puppet army who will doggedly insist that any claims to productivity gains are in fact just hallucinations by people who must not be talented enough developers to know the difference.

It's exhausting.

There are legitimate and nuanced conversations that we should be having! For example, one entirely legitimate critique is that LLMs do not tell LLM users that they are using libraries who are seeking sponsorship. This is something we could be proactive about fixing in a tangible way. Frankly, I'd be thrilled if agents could present a list of projects that we could consider clicking a button to toss a few bucks to. That would be awesome.

But instead, it's just the same tired arguments about how LLMs are only capable of regurgitating what's been scraped and that we're stupid and lazy for trusting them to do anything real.