Hacker News Reader

Comment by tudelo | Hacker News Reader

It is extremely easy to burn tokens if that is required. Explore this codebase. Team x wants y feature, research and generate a full plan. What does feature x in codebase y actually mean? Analyze code coverage in x. Map out code flow and find concurrency bugs in y and on and on...

Oh and my favorite: Use 5 independent subagents to review code change and summarize the findings, and for any finding determine if they are real concerns

View on HN

The other day claude spun up 100 agents and took an hour to type 30k token document to tell me something was impossible to do. I googled it, found a pr on the 3rd link that showed it was possible. "You're absolutely right!!"

"You can't use reflection if the classes aren't in the class loader" "I see why you would think that however this should work, let's test it."

-Claude, burning my company's money.

> Claude, burning my company's money.

And the planet... While I experience some schadenfreude when reading these comments from programmers, I also can not help to wonder when this insanity will this end.

> I also can not help to wonder when this insanity will this end.

When AI use starts to be a line item cost on public companies' financial reports + Anthropic and OpenAI have IPOed and have to file financials too + they kill their growth-hack monthly all-you-can-eat plans.

The entire house of cards falls down when the success metric shifts from "Are you using AI?" to "What return value are you getting for the money you're spending on AI?"

Some smart companies / departments are going to be able to demonstrate stellar AI ROI, but I'm going to be shocked if the bulk of current demand isn't revealed to be naked. Mostly because middle management is always stupid about adopting and using new technology.

There's nothing people run out of faster than other people's money. I expect this second half of the year we see that the cracks in the AI business grow and bring the whole thing down.

Just a bit after anthropic and openAI unload the "value" of their companies into retail investors.

Worse'n crypto… which I would not have believed possible.

> I also can not help to wonder when this insanity will this end.

AI companies are running out of money to subsidize those queries, and worse are needing to show a profit. All while they are having a harder time to raise more money as investment.

If nothing changes, things should become more rational soon.

... now... IMO, I place the odds of nothing changing very low...

I am of a particular disposition that makes it difficult for me to do lie about my work, especially if I were in a firm where people are reading my chat transcripts. As much as I want to stick it to The Man, it feels a lot better for me to just say "no" and burn through my 401k until this blows over.

I'm not too proud to admit that this whole thing scares me though. I fail to see how anything will get better.

Congrats, you have a moral compass and you sound raised by good parents. The other people in these threads bragging about burning tokens, not so much.

I’ve sadly had the same thoughts lately to cash out my 401k and say farewell to software development. I’m hanging in for a little longer, I think the AI/greed fever breaks sometime soon (months not years).

There is value in doing all that too, though. Admittedly with strong diminishing returns, but it's there.

Eg by doing that I was able to develop non-essential features which increased our quality of life for devs last month without going through our PO who'd need to price it - because that does let's you create changes in an incredibly hands off manner with miniscule amount of time investment if you already know what you want to achieve, and how the end result should be...

Admittedly, that's a pretty narrow usecase which is rarely the case- but if it is...

Just ask it to "use a workflow" and it'll spin un dozens of agents burning your token allowance in parallel.

And the more uselessly amusing thing is that the manager who requests higher tokens usage probably also doesn't care whether it's producing slop or not. Metric goes up; managers happy until CFO is reported income hasn't gone up as quickly as costs, and that makes the CEO optimistically concerned. Never expect underlying thought from a messenger.

It's interesting that LLM barely had any vetting period or experimentation phase. Suddenly everyone was supposed to test it in production, it seems.

Let us not forget /ralph-loop “explore the codebase for bugs, write tests for each bug found but do not fix the bug, only capture its existence in testing” will ensure your agent never stops burning tokens.

Afterwards, give me 5 separate documents with 10 plans each for how to implement this. Triple check your work, make no mistakes. Then give me 3 distinct executive summaries emphasizing different areas.

Reply on news.ycombinator.com