Andrej Karpathy 11/17/2025

Verifiability

Read Original

The article analyzes AI as a fundamental shift in computing, comparing it to historical precedents. It introduces the concept of Software 1.0 (hand-written programs for specifiable tasks) versus Software 2.0 (AI-trained programs). It argues that in this new paradigm, 'verifiability'—the ability to efficiently test and reward an AI's attempts—is the key predictor for which tasks and jobs will be automated, leading to uneven progress across different domains.

Verifiability

Comments

No comments yet

Be the first to share your thoughts!

Browser Extension

Get instant access to AllDevBlogs from your browser

Top of the Week