miniblog.

I've had good results by prompting an LLM "review your changes" in the same session when I don't like the initial output. I'm surprised this is effective: I would think it's redundant when you're running with a high effort setting.

I'm adding uptime data to my personal LLM bot. Does this information belong in the system prompt? That was my initial plan, but I'm thinking that a tool would be better. Tools let me see information provenance ("queried the uptime tool"). There's also a small caching benefit.

Anthropic used Claude to write a C compiler in Rust (CCC). I've really enjoyed following the discussion. The initial post shows how they wrote the harness with GCC as an oracle and a prompt specifying a design using SSA. The author has clear domain expertise. https://www.anthropic.com/engineering/building-c-compiler