Reading comments, I’m curios why would someone spend thousands for LLM coding? What are you building to justify these skyrocketing token consumption?

I’ve been using AI coding since GitHub copilot was in beta, used all IDEs in the market, and had very few occasions when I passed the $20 subscription limit. And when I did, that was when I decided to move from cursor to CC and Codex, and still, using them everyday and didn’t have to go above my limits.

▲

jjmarr 3 days ago | parent | next [-]

I've spent US$16 700 last month. I made an autoscaling K8s cluster for distributed compilation/caching on a large C++ project. I also heavily modified the build system to use a forked version of `siso` compatible with our environment.

That meant we can go from 17 minutes on 32 cores to 5 minutes on a few hundred. And because it's distributed compilation we don't have to provision each developer with an overpowered build system they won't be using most of the time.

It could also eliminate our CI backlog because autoscaling. Over a few hundred engineers building this codebase this probably a few thousand hours of waiting a week.

This took me about 2 weeks as someone who graduated 9 months ago. Most of the tokens were spent in several hour long debugging sessions relating to distributed systems networking and tracing through gRPC logs because the system wasn't working until it did.

I think I'd need several years of experience and 6 months as a full time engineer to have accomplished the same thing pre-AI.

Since I work at a semiconductor company near Toronto there's nobody around with the distributed systems experience to mentor me. I did it mostly on my own as a side project because I read a blog post. I literally wouldn't have been able to complete this without AI.

I'm sure the actual solution is terrible compared to what a senior developer with experience would've created. But my company feels like it's getting ROI on the token spend so far even though it's double my salary.

▲

teaearlgraycold 3 days ago | parent | prev | next [-]

I think if you’re a professional and you’re actually coding for >4 hours per day it makes sense. Also if you’re one of those weirdos that likes to command an army of agents.

▲

Oras 3 days ago | parent [-]

Well I’m a software engineer and code more than 4 hours per day.

But I do check the generated code, make sure it doesn’t go banana. I wouldn’t do multiple features at the same time as I have no idea how people are checking the output after that.

I like AI coding and it accelerated my work, but I wouldn’t trust their output blindly

▲

bodhiJhawken 13 hours ago | parent [-]

This is my thoughts exactly.

It speeds up so much of what I do, simple tasks are amazing to delegate to AI and it leaves me with more time in the day to tackle the big tasks.

My general rule of thumb now is I never get AI to build the bones of something which I believe I will need to build something else on top of later. But if it’s some throw away dead end feature that won’t require me to build more tooling on top of in future I’ll happily spin up a cloud agent and use the result.

	▲	teaearlgraycold 7 hours ago \| parent [-]
		They're really amazing for making one-off tools where you just need some clean output and can throw away the code afterwards. I had Claude Opus put together a data labeling web app with very little effort. Less work than creating an account on some SaaS and learning their system.

▲

acron0 3 days ago | parent | prev [-]

I've heard rhetoric like "we have to use LLMs to stay competitive now" which attempts to justify the cost

	▲	sunaookami 3 days ago \| parent [-]
		Reminds me of these "if you don't pay 200 dollars per month for AI you are NGMI" comments...