AI

METR's Time Horizon 1.1 benchmark was essentially saturated by the most capable agents evaluated in early 2026, and building replacements costs over a million dollars in human baselining alone. This course is about the evaluation crisis: why benchmarks are being beaten faster than they can be built, why that makes uncertainty go up rather than down, and how to measure capability when the scoreboard breaks.

Lesson 1 is free. Subscribe to unlock every course.

Secured checkout

$7.65/mo

cancel anytime

Every course on the platform, 350+ and growing

Quizzes, certificates, and your rank on the leaderboard

New courses added every week at no extra cost

Already have All Access? This is included, no need to subscribe separately.

Subscribe

Have a promo code?

USDC & USDT