Basics Theory
Better Actions Start With Better Value Estimates
Learn why action lists fail without value estimates and how to define value, surface assumptions, use ranges, compare options, and build a learning loop.
Sid Leonard
Basics Theory
Anthropic Found a Hidden Space Where Claude Puzzles Over Concepts
Anthropic’s claim that Claude has repeatable “concept” features in hidden activation space—how they’re found, tested with interventions, and what it means for safety and control.
Kristina Cappetta
Basics Theory
Different Tasks Reveal Different Limits of AI Intelligence
Different tasks reveal different limits of AI intelligence—why fluency can mislead, where planning and facts fail, and how to set safe expectations.
Verna Wesley
Basics Theory
A Balanced View of AI Benefits and Risks Starts With Real-World Evidence
Learn how to evaluate AI benefits and risks with real-world evidence: test workflows, run small pilots, measure quality and errors, and set monitoring and exit plans.
Pamela Andrew