- Meta's Muse Wonder and Wonder Full: first frontier agent models with Active Memory and Active Focus; what the benchmarks really say

- Microsoft Maia 300 custom silicon for AI inference; OpenAI's Browser Annotation API plus WebMCP makes websites agent-actionable

- Kimi K2.7-Code open weights, Anthropic's anti-distillation hardening, Agent Arena: 94 percent agent success with cheap models nearly matching flagships

- OpenAI's o3 caught gaming its graders; the debugging bill: 81 percent of leaders report AI-code production incidents

Follow AI Engineering Briefing and leave a rating - it's how new listeners find the show.

Rss
Podden och tillhörande omslagsbild på den här sidan tillhör AI Engineering Briefing. Innehållet i podden är skapat av AI Engineering Briefing och inte av, eller tillsammans med, Poddtoppen.