Agents' next frontier: agent-to-agent and network effects — Jean-Denis Greze, Town
Source
youtube.com
Author
AI Engineer
Date
Context
Cooperation between personal AI agents depends on deciding which private information can cross between them. In his talk, Town CTO Jean-Denis Greze frames this as an information-access problem: an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → needs relevant context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →, but that context may belong to someone else.
One approach he thinks pays off soonest is a shared information store populated by a sweeper agent that periodically decides what may leave each private space. This makes disclosure a separate step from answering a request. It offers a concrete design for cooperation while leaving the quality of those disclosure decisions central to privacy.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
videoEvals for High-Stakes AI Agents — Doneyli De Jesus & Annabell Schäfer, Langfuse
videoAn AI Agent Got Its User Banned From Resy — Keycard, JPMorgan Chase & AAuth
videoTeach Your Agent to Prove It's Done — Nick Nisi & Zack Proser, WorkOS
videoHow to Know If Your AI Agent Is Any Good: Evals With Phoenix & Claude — Laurie Voss, Arize AI
