alt Hacker NewsI am old enough to be retired -- career spent as a research geophysicist. I have about 2-3 scientific projects/papers/codebases left in me before I go be a hermit in a cave somewhere.
LLMs appear to be able help me get through that. I'll take the help.
BUT JEEZ! Dealing with "Rain Man"[1] LLMs week-in and week-out is wildly annoying, frustrating, and anger inducing.
If they all would just STOP sounding so authoritative when spewing nonsense!
[1] After the Dustin Hoffman character in the movie of the same name. "Ah but you know I'm an excellent d̴r̴i̴v̴e̴r̴ programmer."
Not sure what harness you use, but I've found you can influence LLM "voice" a lot more by replacing system prompts rather than just adding content to a CLAUDE.md file. I see a lot of people complaining about how Claude "talks" but very little talking about how to adjust it.
https://code.claude.com/docs/en/agent-sdk/modifying-system-p...
The main benefits to modifying the system prompt or output style is that the harness "reminds" the model about those things every turn. So if you've experienced things like Claude Code losing track of instruction from CLAUDE.md in large contexts, it indicates you may need to add the instruction to the system prompt instead which is far more durable.