This framework depends on specific, evidence-backed contributions. The most valuable submissions identify a data problem, specification weakness, mapping error, or stronger opposing argument that changes what the framework should publish.
"Your series X from publisher Y is wrong / stale / uses the wrong methodology." Open an issue; include publisher citation and why the current series is inappropriate. If accepted, we add a new fetcher or switch the series and re-run affected hypotheses.
"Your hypothesis Z is specified in a way that biases the result." Open an issue or PR; include an alternative specification with justification. If the alternative is reasonable, we add it as an additional spec (version: 2) and publish both. If the alternative changes the conclusion, we update the hypothesis prominently.
"Your hypothesis Z is actually falsified under a correct reading of the data." Submit a PR that runs an alternative specification and demonstrates the falsification. If the falsification holds, we update the result card, publish the change in /updates, and credit the contributor.
"Your steelman file for hypothesis Z does not steelman the strongest opposing argument." Submit a PR with a stronger version. Accepted steelmans credit the contributor.
"Your condition entry for X omits institution Y or misclassifies case Z." Submit a PR with supporting cases and disconfirming cases. Taxonomy entries are living documents.
"Your invariant #N is wrong / impossible to enforce / too strict." Open an issue. Methodological changes require explicit commit messages and, once the review infrastructure is live, independent review.
Not yet paying out (as of 2026-07-18). Successful challenges that change a coefficient, overturn a falsification, or identify a methodological error are planned to be paid in USD when a funded ledger is published in review/bounties/README.md. Until that ledger lists open amounts and a payout process, treat bounties as roadmap — contributions are still welcome and credited.
Contributions and integrity incidents are logged in review/log/ with reasoning (seeded with retrospective internal audits). Transparency is the mechanism. A contribution that is rejected should receive a logged reason. External challenge volume is still early; the log is real but not yet a high-throughput programme.
review/submissions/<date>_<short-name>.md describing the submission and the evidence or reasoning behind it.review/log/.Contributors may include relevant context about their expertise, priors, or exposure when it helps readers interpret the submission. This is optional. The contribution is judged on its evidence, code, and reasoning.
Engage with the strongest version of the position you disagree with. Write for the public review log, not for point-scoring. The framework's credibility is the shared asset; undermining its reviewer pool by hostile posture hurts everyone using the system.