Publish every study
What we do. Every study goes public when it is released, including the ones that fail. Each opens with four answers: what we tested, what we expected, what happened, and what it forces us to change. Then the full report: sealed design, results, post-mortem, audit rounds.
What you can check. That no study is missing. The ledger lists every study in order. That failures get the same treatment: the same layout and the same place in the ledger as a study that works.
Now. Sealed. None published yet.
Sealed designs
What we do. Before a study runs, its design is fixed: what is tested, how, and what we expect to see. The design document is published in full with the study.
What you can check. That the results answer the question the design asked, not one chosen afterward. Read the sealed design next to the results.
Now. Each design is published with its study.
Adversarial audit
What we do. An independent auditor, with no ties to the maintainers, tries to break each study. Every objection, every response and every round goes into the audit trail.
What you can check. Which objections were raised, how each was answered, and which ones changed the study. Objections that were not resolved are marked as unresolved.
Now. The audit trail opens with the first study.
Paper first
What we do. The system runs on paper, in public, before any real money. Fills are modeled, and the assumptions behind them are published.
What you can check. The assumptions, and each week’s comparison of what the model expected with what the paper book recorded.
Now. Not started. Follows the studies.
Then real money
What we do. The book flips to real money. Positions are attestable on-chain.
What you can check. That the positions exist, without taking our word for it. How to verify them is published when the book goes live.
Now. Not started. Follows the paper book.
Drift, every week
What we do. Every week, we publish the realized result next to the modeled result, and the drift between them. When the drift is large, we publish why.
What you can check. Whether the model describes what happened. Every figure re-derives from public data.
Now. Starts with the paper book.
What we don’t publish
- No target rate.
- No estimate or range before a study is released.
- No result without its source.
Read next: Research · Roadmap · Why we publish the failures