Anthropic Discovers Claude's J-space
ZabihullahAtal · x · 2026-07-20
Anthropic's latest research indicates that Claude forms an internal representation during reasoning, which researchers call "J-space".
This mechanism isn't explicitly manually designed but emerges naturally during training. Researchers liken it to a shared whiteboard inside the model: critical information is centralized here before being used for synthesis, planning, and multi-step reasoning.
Experiments show that if J-space is disrupted, Claude can still output fluent text and remember facts, but its abilities in multi-step reasoning, planning, summarization, and complex problem-solving significantly degrade. The research concludes by emphasizing that modern AI intelligence relies not just on "next-token prediction," but also on how internal information is organized and integrated.
More from Models
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11
- 105 hidden bugs, 2 repos: DeepSeek V4.1 Flash fixes 24 at $1.80 vs Opus 5's 27 at $51.33 — ChartsJournalX · 2026-09-11
- awesome-llm-leaderboards: an open-source directory of LLM leaderboards, pricing tables, comparison tools — Last_Establishment_1 · 2026-09-11
- Anthropic claims it works to keep eval environments unidentifiable to models — MaxKannen · 2026-09-11
- Nex N2.5 Pro released on Hugging Face with 407GB of weights — jinnyjuice · 2026-09-11
- RoMa v2 image matching model unveiled in the usual black poster — ducha_aiki · 2026-09-11