The components of agent memory
For Claude to learn from experience, it has to decide what’s worth remembering before it knows what’s next. Here’s how “dreaming” helps it organize memory and correct mistakes.
Sources
Around the web
- The Components of Agent Memory Hacker News · 145 points · 73 comments
- The Components of Agent Memory Hacker News · 7 points
On Bluesky
Anthropic and OpenAI have both landed on 'dreaming,' background processing that lets AI clean up and reorganize its memory between sessions, as the key to assistants that actually learn over time. www.anthropic.com/research/com...
David Hartel (@davidhartel.bsky.social) Nov 7, 2026 · 3 likesAnthropic Engineering: agent memory "dreaming" = offline reflection on transcripts to fix mistakes, write notes, build wiki-style memory, promote into future sessions. Evals: CL-Bench, Pokémon Red, doc wikis. https://www.anthropic.com/research/components-of-agent-memory
Antone Heyward (@antoneheyward.bsky.social) Nov 7, 2026The Components of Agent Memory #HackerNews https://www.anthropic.com/research/components-of-agent-memory
HackerNewsTop5 (@hackernewstop5.bsky.social) Nov 7, 2026Anthropic on Claude agent memory: a "dreaming" pass between sessions can correct notes against transcripts, build wiki-style maps, and promote a small layer into new sessions. Maker claims in the post. www.anthropic.com/research/co...
Kol Tregaskes (@koltregaskes.bsky.social) Nov 7, 2026Claude got lost in Pokémon after saving the wrong location for a PokéCenter. In Anthropic's test, an offline 'dreaming' pass checked that note against its transcript and fixed it. The run then reached Cerulean City. https://www.anthropic.com/research/components-of-agent-memory
@admeshai.bsky.social Nov 7, 2026