OPEN PROTOCOLS / 03
Benchmarks,
not claims.
Every benchmark fixes the model, tools, workspace, time, and rough brief. It removes the operator playbook and counts only independently verified outcomes.
Step-away benchmark
How long useful work continues after the operator stops prompting.
Company-system benchmark
How much of a mature software company an independent verifier can prove.
Fulfillment benchmark
How long a rough idea takes to reach funding and 95% shipped.