Latest / Frontier models and safety
Anthropic tests how Claude writes, checks and organises agent memory
See how this story connectsAnthropic’s engineering team describes how Claude forms memory notes for agents, including a process it calls dreaming, in which Claude reflects on logs to correct mistakes and write new notes. In a Pokémon Red run, Sonnet 3.7 wrote 413 memory files across 225 folders and earned one badge, while Claude Fable 5 wrote nine files in two folders and earned four badges. On the CL-Bench 1.0 database task, the post says Claude Opus 5 was the first of the models it compared to close the learning loop, with most of its entries holding a rule, answers to earlier questions and warnings about traps.
Why it matters
Teams building agents on Claude can use dreaming to test memory notes against the original transcripts before a later session reads and trusts them.
- When
- 6 Nov 2026
- ID
WTA-N-2026-0785- Where
- International
- Who
- Anthropic
- Topics
- AI agents, Frontier models
Summary written with AI by wheresthe.ai, from Anthropic. Checked against the source on 6 Nov 2026.
Source copy: saved on 6 Nov 2026 to check this story and not republished, because the text belongs to its publisher. Also held by the Internet Archive.
Cite this story
WTA-N-2026-0785: “Anthropic tests how Claude writes, checks and organises agent memory.” wheresthe.ai, 6 Nov 2026. https://wheresthe.ai/d/intl-anthropic-tests-how-claude-writes-checks-and-organises-a-2026-11/ (source: Anthropic).
Spotted an error? Email [email protected].