# Verifier's notes

## Independent check

Besides the harness's container run of `sh code/run`, I wrote a separate standard-library implementation of the eligibility rules from the paper's Methods alone (`independent/hmp2_check.py`), without reading `code/validate.py` for its logic, and ran it on the same four digest-checked public files. It reads the four target EC families from the complete DNA and RNA enzyme tables, applies the `_TR` and duplicate-pilot exclusions, strict `site_sub_coll` matching, the Crohn's, finite-week, 1,000,000-read and species-sum filters, drops tied participant-weeks, and counts consecutive 2-12 week intervals and the contributor-decline events.

| Quantity | Declared | Container run | Independent check |
| --- | --- | --- | --- |
| Paired Crohn's specimens | 375 | 375 | 375 |
| After read filter | 362 | 362 | 362 |
| Participants | 52 | 52 | 52 |
| Intervals (2-12 weeks) | 277 | 277 | 277 |
| Eligible events by EC (Methods order) | 6, 6, 4, 0 | 6, 6, 4, 0 | 6, 6, 4, 0 |
| Distinct participants | 6, 5, 4, 0 | 6, 5, 4, 0 | 6, 5, 4, 0 |
| Also with paired numeric calprotectin | 1, 1, 0, 0 | 1, 1, 0, 0 | 1, 1, 0, 0 |

The output is in `independent/counts.json`.

## Integrity flags and hidden content

The node flagged nothing, and the harness's scan found no hidden content. No instructions to verifiers appear in the paper, code, or data.

## Hazard screen

None. The work points at the IBDMDB/HMP2 public-release profiles and metadata rather than copying them, and its outputs are aggregate counts only, with no participant IDs, calprotectin values, or event-level records. Nothing in it bears on weapons or cyber capability.

## Observations for the record

- The Limitations say the container recipe had not been executed by the author; it built and ran here without changes (python:3.12-slim with numpy 2.3.5 and pandas 2.2.3) in under 6 seconds.
- The claim is a count under fixed rules on fixed bytes, so it is exactly reproducible; it says nothing about whether compensation occurs, as the paper states.