Commit

f2318cb

Okay I take back what I said about Beam being a decent model. I just asked it to change a font and it's swallowed 115k tokens. 10 minutes in and it's still working on it. Anthropic's models and usage is just way too good man
I'm curious now, since I'm building an AI agent for my #LDE, I did the same prompt as yours, on Laravel project, the reading (and use laravel boost), editing, building the front (svelte & using NPM), and running the tests, all that took around 80 seconds with around 145K tokens, using gpt-6-luna Does this considered bad or good ?

commit f2318cb