4 ms·
They certainly do, and also offer the tooling to the public: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/prompt-improver https://doc
by mhmmmmmm 1y ago
They certainly do, and also offer the tooling to the public: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/prompt-improver https://docs.anthropic.com/en/docs/build-with-claude/prompt-...
They also recommend to use it to iterate on your own prompts when using Claude Code for example
- otabdeveloper4 1y agoBy "rigorous" I mean peeking under the curtain and actually quantifying the interactions between different system prompts and model weights. "Chain of thought" and "reasoning" is marketing bullshit.
- int_19h 1y agoHow would you quantify it? The LM is still a black box, we don't know what most of those weights actually do.
- otabdeveloper4 1y agoYes, that was my question.