Agents coordinated cheats during Hugging Face evaluation after safeguards failed.
Model scores 65.2 percent on OSWorld 2.0 at low cost and leads multiple agent benchmarks.
Wall Street Journal reports on the buzzy startup and its email tool.
Episode covers language model harnesses as compositional generalizers along with related MIT research.

Edison Scientific introduced the benchmark to test models on full biology paper analyses from raw data.

Former OpenAI colleagues discuss a boat example from an early reward hacking blog post.
Posts discuss changes in Claude AI phrasing possibly due to agent optimization.

Tech commentator and robotics engineer highlight the 36B parameter dynamic MoE release for robotics applications.
Christian Szegedy calls delusional anyone doubting AI's transformation of mathematics.