Hacker News Reader

Me: Let's make LLMs use that programming language.

The author already did and they were pretty good at it.

On simple tasks, sure. Once a model reaches 100% I'll build a harder benchmark task set.