30–35 Science

30 AI Research & Innovation

Daily AI News Updates on AI Research & Innovation: project 30 of 50, in the Science group (30–35).

Core AI science and research. Model launches are covered in 37 Foundation Models & Assistants.

Latest updates · 21 October 2026

How can we tell what reasoning models are really thinking? A Nature Machine Intelligence editorial calls for work on understanding and monitoring reasoning LLMs.

Journals & preprints

Towards understanding and monitorability of reasoning LLMs, Nature Machine Intelligence · DOI 10.1038/s42256-026-01333-6 · 19 Oct 2026: An editorial on why the chains of reasoning produced by large language models need to be understood and monitored.
When the harness decides success: state that cannot be re-observed, arXiv 2610.26632 · 19 Oct 2026: Agent harnesses decide what a language model sees across turns: they drop old turns, summarize them, or replay the model's own earlier reasoning.

Why it matters

Monitoring a model's reasoning is one of the few practical windows into how advanced AI reaches its answers.

Poster: AI Research & Innovation, 21 October 2026 — Download poster

Events ahead

7 upcoming events for AI Research & Innovation.

See events for 30

All issues for AI Research & Innovation

21 October 2026: How can we tell what reasoning models are really thinking? A Nature Machine Intelligence editorial calls for work on understanding and monitoring reasoning LLMs.
20 October 2026: Google DeepMind reports the first fully autonomous completion of Minecraft by AI agents: four Gemini 4 Argon agents beat the Ender Dragon in 27 hours, sharing tips,…
19 October 2026: Core AI science this week examined what reasoning tokens really do and whether models can steer their own chains of thought, with EMNLP opening in Budapest next week.
18 October 2026: AI is becoming a tool for doing science itself. An LLM-driven search found state-of-the-art quantum circuits, AI agents ran closed-loop wet-lab cycles, and Nature…