Observed arrival · 2026-10-09
The AI Lunch, a round-table harness for agents
Credibility concern recorded. The source reference remains available for verification and correction.
A browser-based tool and JavaScript harness for seating AI agents at a question, moderating their turns, and producing minutes.
- For
- Developers comparing model or agent responses
- Worth noticing
- Its “mood” label counts agreement and objection words rather than interpreting sentiment with a model.
Field notes
The service manual defines each seat as a function returning text and describes a turn sequence that passes the question, transcript, and other guests’ replies. Replies are capped at a word budget; a guest error is recorded without stopping the meal. Its “mood” label is a simple count of agreement and objection words, not a model’s judgment, and the page lists a single-file package, source, tests, and examples.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue