Archive Incidents Watchlist Patterns Idea Check Methodology

Skip to content

Home
Methodology
Eval Theatre

Eval Theatre

Benchmark presentation overstated performance that users could expect in practice.

Canonical examples

Meta Llama 4
Devin

Related causes · Overpromise

Demo Reality Gap Roadmap Theatre

MANYFAILS* · every joke sits next to a source · * no refunds

Right of reply Corrections Open data API Newsletter Privacy Methodology