Reaction
Systems thinking and evals for building with AI agents
A post recommends examining a system’s subparts and checking data and traces to judge progress.
TLDR
One commenter says thinking about a system’s subparts helps people work effectively with AI agents. They also recommend measuring outcomes and building evals early. Agents can help people get started, they say, but data and traces are needed to judge whether the work is making real progress.
Combined views
2.4K
1 Source, first seen ago