The rules, in full

Anyone. No PhD, no affiliation, no lab. What is required is one piece of prior work — a paper, a repository, a writeup, a thesis chapter — that shows you have taken something from a question to a result. It is reviewed once and carries across cycles.

An empty affiliation is not a mark against you. People outside institutions are a large part of who this exists for.

A question, a method, and the result that would kill the project. The last one is not optional and not decorative: it is stored with a minimum length, reviewers read it before anything else, and a falsifier that turns out to be a risk rather than a result is the most common reason a proposal is declined.

“We might not get enough data” is a risk. “If decoder accuracy on device data stays within 1% of the uniform-noise baseline, the premise is wrong” is a falsifier.

Five dimensions, scored one to five:

• Clarity — can a reader restate the question after one pass?
• Falsifiability — is the kill criterion reachable?
• Feasibility — can a distributed group with no lab do this?
• Stack relevance — does it produce reusable software or data for the quantum stack?
• Lead capability — does the work sample support running it?

Intake is open to anything in quantum. The weighting toward hardware and methods lives in that fourth dimension rather than in a rule about what may be submitted, so an outstanding proposal outside that centre still gets in.

Three applications per person per cycle, enforced by the database rather than by good manners. Applying is free in money, so it has to cost something in attention — otherwise ten good groups drown under three hundred applications each and the thing we promised leads stops being true.

You answer the lead's technical question rather than writing about yourself, and the lead reads that answer before seeing your name and affiliation.

Halfway through, every group is tested against its own kill criterion — the exact text, snapshotted when the group was admitted, so it cannot be softened along the way. The verdict is continue, pivot, or kill.

A group that is killed publishes what it found and closes as completed on criterion, not as a failure. Reaching a clean negative result in eight weeks is a better outcome than eight months of drift, and it is the outcome this place is built to make cheap.

If you want to see what one reads like before writing your own, Cogitan publishes its own closed directions on the same terms — two of them, with the reason each ended and what would reopen it: room-temperature spin registers and learned readout discrimination. Neither was a Commons group — they were our own work, and they are linked here rather than sitting in the record because the record is for groups. Asking people to publish what stopped them is easier after doing it.

We provide structure: collaborators, a cycle with a deadline, a workspace that records what was ruled out, and a public record with your name on it.

We do not provide funding. Cogitan has no budget for this and would rather say so than imply otherwise. There is no lab access and no fridge time either — groups here do software, theory, and data analysis, which is most of what a distributed team can do in any case.