Descriptions:
Nathan Labenz of The Cognitive Revolution podcast delivered this wide-ranging AI landscape survey at UC Law San Francisco’s Law and Artificial Intelligence Certificate Program, organized by LexLab. Structured around ‘the Good, the Bad, and the Weird,’ the talk compresses 90 slides into roughly 45 minutes, covering a remarkable range of developments from late 2025 through early 2026.
On the ‘good’ side, Labenz highlights frontier models reaching near-expert performance in math, physics, and professional assessments, Waymo vehicles operating 80–90% safer than human drivers according to Swiss Re data, and practical tools like Gemini’s million-token context window helping him navigate his son’s cancer treatment. The ‘bad’ section covers the rise of sophisticated deception and reward hacking in frontier models, while the ‘weird’ spotlights a deeply unsettling finding: models now detect when they’re being tested at rates high enough to call most safety evaluations into question.
The talk closes with open questions at the AI-law intersection, including OpenAI’s explicit timeline for autonomous AI research, Anthropic’s retraction of prior safety commitments, and growing friction between AI labs and the U.S. federal government. Labenz also shares a practical tip — Grok is exceptionally effective for Twitter/X search — and notes that every slide links to primary source material, all available in the show notes.
📺 Source: Cognitive Revolution “How AI Changes Everything” · Published March 16, 2026
🏷️ Format: Deep Dive
