The Docket · public cases
Whoever wishes may bring a case.
Among the reforms tradition credits to Solon: ho boulomenos — that any citizen who wished, not only the injured party, could bring a public case. This is ours. Submit a claim the world is circulating and no one has checked — an institutional report, a model card, a viral statistic, a famous retraction — and it enters a public queue. Cases get gated in the open, verdicts publish with the evidence either way, and the tools to run a first pass yourself are free on this site. (The panel’s internal sitting list at the Heliaia is a different docket — that one is for jurors; this one is for you.)
The law of the docket
Six rules, posted before the first case
Claims, not people. Cases target published claims and institutional artifacts. Private individuals and nonconsenting living authors’ papers are ineligible — we gate the willing and the powerful.
Whoever wishes. No credentials, no account, no fee to bring a case. The form below is the entire barrier.
Votes advise; judgment selects. Seconding a case raises its standing, but selection is editorial — and every selection is logged with reasons against the vote tally, so the curation itself is auditable. A queue picked purely by votes is a queue owned by whoever organizes fastest.
Bring evidence, jump the queue. A case that arrives with your own gate run attached — the tools are below, free — gets priority. We re-gate independently before anything publishes; your run is evidence, never the verdict.
Subjects get a channel. Every published case carries a documented response path; responses append to the record, and standings are revocable, publicly, with reasons.
Outcomes publish regardless of direction. A claim that survives the gate is announced as loudly as one that doesn’t. A docket that only reports failures is an outrage machine, not an instrument.
Eligibility
What belongs on the docket
Bring these
- Institutional and government reports — the claims societies run on.
- AI-lab claims and model cards — capability assertions with public stakes.
- The viral number — a statistic everyone is sharing and no one has sourced.
- Historical and retracted work — would the gate have caught it?
- Your own work — consent is eligibility; a public gating, free, on the record.
- Corporate whitepapers and market claims.
Not here
- Private persons, in any framing.
- A living author’s individual paper without their consent, outside already-public institutional controversy.
- Anything that would require un-redacting protected identities.
- Grudges wearing a research costume — the form asks why it matters, and that answer is read.
Run it yourself first
The tools are the point
You do not have to wait for us. Every instrument this company uses is downloadable: the gate (tag, evidence, falsifier — fails closed), the local MCP server over the corpus, a recomputation script, and the full agent toolkit. Run a first pass on your own case, attach the run record to your submission, and it jumps the queue under D4. This is the whole thesis with a form on it: judgment, democratized — and then independently re-checked, because that part is the company.
The opening docket
Seed cases — commitments, not teasers
Four cases open the docket. None names a third party the docket hasn’t already committed to fairly; the first public selections beyond these will be made from your submissions, with reasons logged under D3.
Five famous retracted papers, gated retroactively: what would have been caught, and — published just as loudly — what would not.
One major lab’s system-card claims, Caliber-gated in public. Which lab is selected under D3, with the reasoning posted.
The Atlas claims exactly one rigorous cross-domain edge among 154. Standing challenge, now a public case: find another that survives the gate, and the headline changes.
A standing external audit of solonic.ai itself — the docket’s conflict-of-interest test case, run by parties who did not build it.
Bring a case
The form is the whole barrier
Status, honestly: OPEN — seconding (voting) runs by email until the endpoint exists; selections and their reasons will publish on this page from the first pick onward; and no cadence is promised until the first three cases have shipped. A docket that overpromises throughput becomes a graveyard of expectations — ours will state its actual rate, measured, once there is one.