XNews

Jason Wei's AI Framework: Intelligence Commoditization, Verifier's Law, and Jagged Edge

Jason Wei's AI Framework: Intelligence Commoditization, Verifier's Law, and Jagged Edge

Jason Wei, a researcher at Meta Superintelligence Labs and former OpenAI contributor, presented three frameworks for understanding AI in 2025: intelligence is becoming a commodity as costs drop; the Verifier's Law states AI capability correlates with task verifiability; and the jagged edge of AI capabilities means progress is uneven across tasks, with some remaining resistant to automation.

Key insights

Intelligence Commoditization and Its Implications

Wei argues that AI progress has two phases: pushing frontiers and commoditization. MMLU costs drop yearly, driven by adaptive computation (o1-style) that allocates compute based on problem difficulty. Access to public information is also becoming near-free, with agents like OpenAI Operator able to navigate databases. This democratizes knowledge but increases the relative value of private internal information.

Verifier's Law: AI Can Solve What It Can Verify

Wei proposes that AI's ability on a task is proportional to how easily that task can be verified. Key factors: objective correct answer, verification speed, parallelizability, noise level, and reward signal continuity. Benchmarks are themselves designed for easy verification, explaining why they are being solved. He recommends AlphaEvolve as an example of exploiting verification asymmetry by iterating on a single problem.

The Jagged Edge of AI Capabilities

Wei rejects the 'fast takeoff' hypothesis, arguing AI self-improvement is gradual and task-specific. He provides three heuristics: AI excels at digital tasks with fast iteration; tasks easy for humans are often easy for AI, except those humans physically cannot do (e.g., analyzing millions of medical images); data-rich tasks perform better. He predicts AI research automation by 2027, but physical tasks like plumbing and haircuts remain distant.

Source posts