Research

  • [OpenAI] GPT-5.6 Sol, reported to beat Mythos on TerminalBench.
  • [arXiv] Binary questions are better for LLM-as-a-judge setups.
  • [Simon Willison] More discussions about the new GLM.

Business

  • [OpenAI] OpenAI’s inference chip.
  • [Tweet] OpenAI’s keyboard hardware for Codex.
  • [Claude] Claude with its own account.
  • [Tweet] Pricing concerns in industry adoption.
  • [Polsia] Polsia - company as a service?

Other

  • [levels.fyi] Interactive map of the bay area with levels.fyi data.

Have a great week!

Updated:

Comments