Vibeleaderboard
← All Intel
Intel / video

Agents that remember

Source
youtube.com
Author
Claude
Date

Context

By default, Claude agents on Anthropic's Managed Agents platform start each session knowing nothing about earlier ones. In a workshop, an Anthropic engineer named Kevin shows the fix: a memory store, a file-system-like store attached to a session, which the reads and writes with ordinary tools such as search. In the demo, facts given in one session were found and used in a later session attached to the same store.

Stores can grow messy, so Anthropic added dreaming, a background job that takes a memory store plus transcripts of past sessions. A process, with one per transcript, checks facts, adds details such as dates and identifiers, merges duplicates and writes a new output store, leaving the original untouched; the console shows a diff, so a person can review what changed. The demo was one small example, and the speaker describes benefits such as easier retrieval as hopes, not measured results.

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.
  • multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
  • subagent — A separate agent spawned by another to do one scoped piece of work in its own context, returning only the result.
Read the source www.youtube.com
More from Claude
Recommended reads
Comments

Checking sign-in…

Loading comments…