A safety institute argues capability is a curve, not a number

Claim

The UK's safety institute countered that our rulers are too short, since capability is a curve over test-time compute: one model's cyber time horizon stretched from two hours to fourteen as its budget rose from 2.5 million to 50 million tokens.

Reported metric: 2 hrs to 14 hrs

Reporting date: 2026-07-03 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to July 3, 2026 (theinnermostloop.substack.com)