20 October 2026
Mollick charts AI progress over 30 months, citing task-length and cost gains
First reported
One Useful Thing ran this on .
- Ethan Mollick, a Wharton professor who writes about AI, tracks how far the technology has advanced since his first book.
- A benchmark called METR, which measures how long a task an AI can finish on its own, rose from about 4 minutes in 2024 to roughly 17 hours by April 2026.
- Models at the level of GPT-4, an earlier OpenAI system, now run on phones, and the cost of using AI has fallen.
Our read
Only one of the 30 tracked newsletters, One Useful Thing, covered this story, first reporting it on 2026-10-20. The archive's 2026-10-21 Mollick story on AI improving fast but unevenly is the closest earlier piece on this thread.
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
How it was covered
One Useful ThingEthan Mollick
Book launch, METR benchmark trends, model leadership changes, falling prices, a benchmark cost comparison, an AI-generated otter video, and how to work with agents