Adoption lags frontier: why OpenAI staff AI use may mirror average users

maksym_andr · x · 2026-09-26

In a debate over OpenAI's internal model adoption, the author argues adoption is not a direct measure of frontier capability: new models take time to pass internal testing before being deployed inside frontier labs to accelerate their own R&D, and public deployment takes even longer.

The counterpart notes the median OpenAI employee isn't as AI-pilled as assumed — using AI about as much as the average Twitter user, possibly less — which matters for calibration. The author hints one benchmark may reflect unreleased model capabilities.

Related event: Researchers: Measure Model Gaps by Capability, Not Months(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →