Computer use has made much slower progress than other verifiable domains like coding and math, despite being clearly verifiable (e.g., 'did the package get delivered?', 'is the venue booked?', 'have taxes been submitted?'), which is unusual and reveals something important about AI progress constraints.

factualpending

Speaker

Unidentified Speaker — What does the next training paradigm look like? [20p5-kQXF_Q]

Evidence Quote

Why has progress on computer use been so much slower than other domains? Computer use is so clearly verifiable... So isn't it weird that computer use has been making so much slower progress than coding and math

Source

What does the next training paradigm look like?Dwarkesh Patel
Created: 8/12/2026, 6:40:37 PM

My Notes

Loading notes...