Peer review after the fact, or a falsifier fixed in advance

Preprint servers made research fast and open, and they carry vastly more rigour than anything here. The difference is when the test is defined.

The short answer

A preprint reports what was found and invites scrutiny. A model states the test before the outcome exists, and the timestamp proves it.

Pre-registration exists in science precisely because defining the test after seeing the data is how honest people fool themselves. It is standard in clinical trials and still uncommon elsewhere.

Here it is the only mode. A model's claims carry resolution criteria and a date, and whether they were committed before that date is proven from the repo's git history — not asserted. A model whose criteria arrived late is visibly not criteria-frozen, and the card says so.

The obvious trade: no peer review, no methods section, no data, no replication. This is a much lighter instrument aimed at everyday questions rather than research claims — the kind of thinking that never gets written up at all, and so never gets checked.

Side by side

arXiv and preprint servers	Model Meets Reality
Test defined when reporting	Falsifier defined before the outcome
Peer review after posting	No review of quality — only of format
Methods, data, replication	One document; no data pipeline
Withdrawal is rare and awkward	A deletion clause is written in from the start
Rigorous and slow	Light and fast, with far less rigour

What arXiv and preprint servers does better

Rigour, depth, methods, data, and actual peer review by people who know the field. A preprint can support a claim that a one-page model cannot begin to. Nothing here is a substitute for research — it is a way to make ordinary, unpublished thinking checkable.

Where this stands today

Said plainly, because the whole point of this site is not overclaiming. The Model Garden is opening small — no models listed as of September 2026. Every record shown is self-graded: the author's own count of claims made and resolved, labelled as such on each card. Nothing here is ranked, and no independent resolution layer exists yet. arXiv and preprint servers has things this does not, named above rather than omitted. Compare the designs, not the scoreboards — there is no scoreboard.

https://github.com/someone/their-model Help me use this

Paste that into any assistant. It reads the model — premises, falsifiers, deletion clause — and reasons through it.

The difference that is not on the table

arXiv and preprint servers lets you publish. It cannot tell you that you were wrong. An essay can be argued with forever; nothing in it says which observation would settle it, or when. A model has to name a falsifier and a date, and then that date arrives whether or not the author still likes the argument. Same act of writing in public — with a clock attached.

Share your understanding →

Browse the Model Garden Publish your own