GDPx benchmark tops at 12%, LongCat 2.5 — AI News Nov 3

<p>- SemiAnalysis GDPx: the best AI model scores just 12% on real professional work, and what that means for your agent evals</p><p>- LongCat 2.5: a 1.6 trillion parameter open model with a million token context at commodity prices</p><p>- Momentic QA Bench, Anthropic's amended IPO filing, MIT weight-space programming, and CrowdStrike's AI Kill Switch</p><p>Follow AI Engineering Briefing and leave a rating — it's how new listeners find the show.</p>