Hacker News new | ask | show | jobs
by rpunkfu 1 hour ago
Essentially pay-to-win for coding.
2 comments

They even replicated the lootbox logic. How long until the agent gives you a set of changes you have to open with a paid key (60% of success)!
With moves like this, I'm not so sure victory is assured... but yes. Pay to play.

I, like most, beat linting delay by running it in the background; asynchronous.

Having been negative, so far, I want to say something positive. I appreciate that LLMs weren't used for license washing in this case.

That is why rewriting in another language is such a cheat. LLMs beat 10 years+ experts at the details of a single line of code (and I mean SYNTAX, not invariants), but suck at the design, at the level of functions, and higher up.

If you rewrite someone's stack ... you take out the design part of the work. No need to redesign. Obviously the high level decisions work. 99.5% work as well in python as they do in rust ...

Of course, this will never be the next big thing, and just doesn't help companies one iota. If you need a new thing, this will fuck up badly. Hell, even this translation will fuck up without a senior developer regularly telling the model: "WTF!!!!! Don't do that. Just don't". For a new product you need 100x that.