Comment by fxtentacle | Hacker News Reader

Comment by fxtentacle | Hacker News Reader

Parent Full thread

Code is not cheap. It's just heavily subsidised with VC money. But that won't last forever.

Uber Eats also used to be dirt cheap. Surprise! it's not anymore.

And even if you just pay API prices for Opus - as opposed to using a subsidized subscription - you can easily reach the point where the tokens for AI-generated code become comparable in price to just paying a junior dev salary for a manual implementation. AI is great for greenfield projects, where there is little to no existing context. But on real codebases, people memorize large parts of it. That allows them to navigate files with 100k+ tokens in them. (Wherease the Opus API will charge you $2.5 for each time the model runs through 100k thinking tokens reviewing your file.)

But what AI can imitate pretty well is the result of having a clueless middle-manager review your code. So my prediction would be that the AI "revolution" will slim out management layers before it'll reach actual developers.

I doubt that all of the providers hosting open models on open platforms are losing money on serving inference. They have the benefit of not having to pay for training, but the models are open and aren't going away anytime soon.

Sadly, I have not been able to find any open model that comes close to Opus 4.6. So while they are much cheaper to deploy, they also aren't good enough for unsupervised agent execution. But you need a model that can run unsupervised for the claim "Code is Cheap" to become possible.

I don't really think so. Maybe it's because the systems I build need to be reliable and understandable to humans, but I don't think Opus 4.6 is good enough to be unsupervised. I've spent a lot of time using it and I have to tell it no semi frequently and rewrite by hand/iterate with the model frequently. It's saves me a lot of time when used this way, I think, but I still have to give it overall structure and keep it scoped to small changes to prevent it from going down wrong paths and generating tons of unnecessary code (which is how you end up with unmaintainable slop). Less code is pretty much always better and these models made it really easy to ignore that until it's too late. This is on a healthy mix of greenfield and brownfield projects.

To that end, I've actually found Kimi K2.5 to be "good enough" for a lot of that, not quite as good at Opus 4.6, but good enough that it gets me like 80% of the value for a fraction of the cost and more speed.

Hardware has always gotten cheaper every year, and always will. You will be able to run Opus 4.6 tier models locally with junkyard hardware in 2035.

Citation needed. Hardware prices have gone up substantially in the past year.

"global warming is not real because it's really cold today"

i guess my point was that it isn't a given that hardware will always be cheaper, considering that memory prices this past year have gone through the roof.

you may argue that it is an outlier and capacity will increase to meet demand, which it probably will, but who's to say that won't just mean further demand from unlimited money ai companies and then further increased consumer prices?

would opus 4.6 be relevant in 2035 ?

Reply on news.ycombinator.com