c4
A benchmark for hand: LLM agents played Connect 4 against me in a real browser, at 20.7 s and $0.022 a move, and lost every game they finished.
Badges come from GitHub once the author claims this feed.
A benchmark for hand: LLM agents played Connect 4 against me in a real browser, at 20.7 s and $0.022 a move, and lost every game they finished.
I wrote what accrete's tools are allowed to do as one Alchemy Domain model.
Don't make the AI do 500 jobs.
A coding agent runs inside amber, which saves it at every model call.
A model wrote 88 agent plans that wait on a person.
I am learning Cheng Lou's Handwired, where a program is a fixed list of FP32 ops.