Anthropic identifies a global workspace analog in Claude

New research from Anthropic describes a structural division within Claude that mirrors human cognition. Only a small fraction of the model’s internal processes are available for explicit description and reasoning, while the remainder operates outside of conscious access.
Up next
How to bring Opus closer to Fable 5 performance using local logs
Every Claude Code session saves locally as a JSONL file tagged by model. If you used Fable 5 before it was pulled, those traces are on your machine. You can filter them, quantify the model’s behavior, and inject the findings into CLAUDE.md to force Opus to plan like Fable 5.
Related stories
- Claude learned proteins, not just papers about them
- Anthropic bought a biotech with one full-time employee
- A perfect refusal score isn’t a safer model, Anthropic says
- Anthropic touts 94.2% true positives; curl rejected 19 of 34
- Anthropic’s IPO filing says a key safety threshold is met
- 1,000 Claude agents redid the Fermat proof in 28.7 hours
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

