On the estimation and validity of AI time horizons---a statistical look at the METR plot
Drew T. Nguyen, William Fithian · UC Berkeley / arXiv · 8 Oct 2026 · not peer-reviewed
01 · The key finding
The metric reports human completion time for software tasks AI solves in half its attempts. It does not measure reliable runtime without supervision.
An AI test does not say how long AI can work alone
02 · The study
03 · What we take from it
Do not use this time metric as deployment approval. Test your own tasks and define when human intervention is required.
What does this mean for your company? Book a conversation ↗
Sources
2027 Global Digital Trust Insights: C-suite playbook and findings
PwC · PwC · 1 Oct 2026
From Token-Max to Outcome-Max: How You Use AI Determines Its Productivity
Chen Xu, Mengqiao Liu, Beibei Li, Chenyan Xiong · Carnegie Mellon University / arXiv · 5 Oct 2026 · not peer-reviewed