NEW RUNTIME
SignalNov 7, 20261 min read5 sources

Scientific agents now have to hand back checkable artifacts, not just answers

In one minute

Physical Superintelligence's proof of the Aaronson–Ambainis conjecture is checked in Lean, and Axiom released a Lean formalization of the whole Stacks Project. Standard Physics' Reynolds returns a simulation case you can rerun, and ReproBridge's refined artifacts raised experiment reproduction from 76.4% to 92.6%.

If you build research agents, make machine-checkable or rerunnable output the deliverable. A prose claim is not enough.

From the Field Notes:

EditorAndrey Reshetnikov, AI product and AI engineering

  1. Nov 10, 2026Claude Fable 5.2 scored 174 on the Epoch Capabilities Index, beating GPT-6 Astra's prior record of 169.anthropic
  2. Nov 10, 2026Introducing Claude Fable 5.2 on AWSanthropic
  3. Nov 10, 2026llm-anthropic 0.31 adds Claude Fable 5.2 support to the llm command-line toolanthropic
  1. Scientific agents need artifact evidence

Evidence / sources

  1. 1psi.inc/research/aaronson-ambainis
  2. 2axiommath.ai/research/the-stacks-project-formalized
  3. 3cfsg.axiommath.ai/CFSG_errata.pdf
  4. 4standardphysics.com/reynolds
  5. 5arxiv.org/abs/2611.08452
All Signals

Next reads

Read next

Open the graph
  1. 01topicAnthropicExplore the anthropic topic hub.
  2. 02archiveSignalsOpen the signals and the raw-signal archive.
  3. 03sourcesSourcesSee the original sources this archive cites.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract