Writing
- What was required, what was judged, and who says so
- A receipt that names who signed it
- A delivery receipt a skeptic can check
- A baseline two parties can name
- Self-Certification is not verification
- Asking the record
- When the benchmark was broken
- A verifiable receipt for your best-of-N run
- How do you keep an AI agent's decisions honest after the evidence changes?
- How do you verify what an AI agent did without trusting it?