How to challenge a number

1. Name the exact reading

A challenge without a file and a clock cannot be checked. Collect:

1. The dataset id from the citation page (for example ddti or weekly-situation).
2. The URL of the latest or history file you used.
3. The generated_at timestamp printed in that file.
4. The seal.payload_sha256 if the file carries one.

2. Reproduce before you argue

git clone https://github.com/beepboop2025/palimpsest
cd palimpsest
python3 scripts/reproduce_all.py

Exit 0 means the sealed chains and derived scorecards still match. Exit 1 names the check that broke. If the verifier already reports the break, quote that output. You do not need to prove the same break twice.

3. Say which kind of error it is

1. Method error. The formula, window, or join is wrong even though the bytes match. Point at the TUNING POINT comment or the method file.
2. Source error. The upstream public file does not say what Palimpsest recorded. Attach the upstream URL and the fetched hash if you have one.
3. Rendering error. The JSON is right and the HTML is wrong. That is a display bug, not a new measurement.
4. Stale file. The page shows an old generated_at while a newer commit exists. Check collector health before treating staleness as a finding.

4. File it in public

Open a GitHub issue titled Challenge: [dataset id] [generated_at] at beepboop2025/palimpsest. Paste the verifier output. Do not send Palimpsest non-public data, in-country vantage details, or anything that would identify a person inside a censored information space.

What a successful challenge does

The next published reading either corrects the method and says so, or keeps the original number and records why the challenge did not move it. Palimpsest does not silently rewrite a sealed file. If a sealed chain is wrong, the verifier is supposed to fail in public.