Anthropic details agent memory: Claude 'dreaming' pass revisits chats to fix saved notes
HankYeomans · x · 2026-11-08
Anthropic's engineering team published a deep-dive on building agent memory for Claude.
Key mechanism: 'dreaming'
- Inspired by how the human hippocampus replays experiences during sleep to consolidate knowledge, Claude uses idle time to revisit original conversations and correct its saved notes, create new ones, build memory representations, and decide what to preload into future sessions.
- This fixes a real failure mode: a saved mistake can carry straight into the next session.
- The idea builds on sleep-time compute proposed by Letta and UC Berkeley researchers.
The post also walks through evaluations the team used to explore agent memory and how each component contributes to memory organization — a hands-on engineering deep-dive on context engineering.
More from coding & agent
- AI Lab Reportedly Fired Engineer for Using AI Agents to Auto-Merge Non-Prod Code — Yuchenj_UW · 2026-11-08
- MCP tool descriptions silently change — and none of the 4 major AI coding tools warn you — abhiramreddykalluru · 2026-11-08
- A simple prompt coda stops coding agents from constantly asking for your approval — doodlestein · 2026-11-08
- Apple fumbled the OS-level personal agent opportunity, leaving browser-based agents as a workaround — omooretweets · 2026-11-08
- Stanford's Memorilla compresses LLM memory to 16 tokens, 4212x fewer and 27x faster — _AustinCalvert_ · 2026-11-08
- 13 free open-source repos that turn your notes into an agent-ready second brain — nikola_mr64990 · 2026-11-08