# evidence — machine-readable benchmark receipts
Every run lands here with its receipt, trace, court results, and replay
summary. **No fabricated performance claims** (§52): values that were not
measured are `null` in receipts and listed under `not_measured`.
## Index
| `0001-demo-mixed` | demo (synthetic; mixed loading+streaming) | 7 | receipt, trace, markdown, courts, replay summary, pak |
## Honesty contract
`0001-demo-mixed` is generated by the deterministic synthetic workload
(`tools/workload-demo`). It proves the pipeline end-to-end and provides the
first KatraFlow shadow measurements. **No number in it is a claim about any
real game.** Real titles produce receipts through the same pipeline
(`katra-bench` / `katra-prof capture`), and those receipts state their
workload identity in the header.
## Reproducing
```sh
cargo run -p katra-bench -- --out evidence/0001-demo-mixed \
--seed 7 --frames 60 --assets 32 --court all
```
Same seed → structurally identical traces (`trace_determinism` court,
`katra-replay compare`).
## Artifact policy
- `*.katra` traces: kept (small; the semantic record of the run).
- `*.pak`: not committed — deterministic, regenerated from the seed by
`katra-bench` when missing.
- Receipts: append-only; a rerun that changes results creates a new run
directory rather than overwriting history.