# How the lab publishes > The editorial standard for this open notebook: how a piece of Eternity II research goes from unpublished work to a published page. What kind of contribution it is, whether it publishes, at what tier, and where it lives. One shared standard, built to scale to many authors. - Canonical page (with interactive figures/demos): https://eternity2.dev/research/lab/experiments/methodology/ - Updated: 2026-07-21 --- This page is the editorial standard behind everything in the lab. It exists so that a finding reaches a reader as an authored artifact, not a raw note, and so the same rules apply no matter who writes it. ## Published, and unpublished The ground rule everything rests on: - **Unpublished work does not appear here.** Where a researcher keeps their work in progress is their own business. It is simply not in this repository. There is no half-finished draft sitting in the published notebook. - **Published work lives in the lab.** Public, distilled, written for a reader. It got here through review. Publishing happens by pull request. A researcher opens a PR that adds or promotes a page, and the review on that PR is where the decision to publish, and at what tier, is made. Nothing enters the public record without a PR, and the PR is where a second reader checks it against the standard below. This keeps review collegial rather than a heavyweight gate, and it scales to many authors. ## Three axes describe every page Keeping these separate is the whole method. It is easy to blur "polished," "reproducible," and "reviewed" into one vague notion of "done." They are not the same thing. 1. **Contribution:** what kind of result is this? 2. **Tier:** how far through review has it come? 3. **Rigor and reproducibility:** how firmly is the claim backed? Plus attribution: who did what. ## Axis 1: contribution Not every result is a solver, and the notebook stopped pretending otherwise. A page declares its contribution: - **solver:** produces a competitive board by searching. Its score is a real search result. Only these earn a row on the leaderboard. - **analysis:** proves or computes a property of an existing board or of the instance. It does not produce a new board. - **reconstruction:** decodes or rebuilds the community's known work to extract an insight. The number is theirs, decoded. - **theory:** a mathematical property, law, or impossibility proof. - **method:** a technique described for reuse, not a scored run. - **measurement:** a benchmark or empirical observation about solvers or instances. - **negative:** a rigorously-run dead end. First-class, not a footnote. - **tool** and **exposition:** a software artifact, or an explainer. The load-bearing distinction is solver versus everything else. A meet-in-the- middle band solver that proves an endgame is optimal is an analysis, not a solver, and it does not belong on a score chart even though it emits a number. Getting this axis right is what lets the leaderboard mean one thing. This axis is also a browse surface: the [by-contribution index](/research/lab/experiments/by-contribution) gathers the pages by the label they declare, so the analyses, the theory, the measurements, and the [dead ends kept as first-class negatives](/research/lab/experiments/by-contribution/negative) each have a shelf of their own. ## Axis 2: tier Once a page is published, it carries a tier that says how firmly it has been reviewed: - **Technical report:** public, but the review confirmed only that it is sound and fairly stated, not that it has been independently checked. The page carries a "technical report" badge, the way a preprint is stamped "not yet reviewed." A legitimate resting state: not everything needs to go further. - **Reviewed finding:** a second reader, or the same author after a cooling-off, confirmed it holds. Cited, treated as close to immutable; corrections happen in place, annotated. Two principles carry over from how research publishing works elsewhere. The tier is stamped on the page, so a reader always knows what they are looking at. And promotion turns on rigor, not on outcome: a sound method that found nothing gets published as a negative result, while a striking result on an unsound method does not. ## Axis 3: rigor and reproducibility Every page states how firmly its central claim is established (proven, measured, or conjectured) and how re-runnable it is. For anything quantitative the gate is simple: a number publishes only when its exact configuration and seed are archived and it can be re-run from the documentation. A benchmark score with no re-runnable config does not get published. Cheaper contributions carry a lighter bar: a negative result or an explainer needs sound reasoning and clearly stated limits, not a reproduction script. ## Attribution, built for more authors Each researcher's pages gather automatically on their contributor page, derived from the byline, never hand-listed. Adding an author is adding an entry to the registry and writing pages under their name. When a page has more than one hand, a light contributor role (who ran it, who analysed it, who validated it, who wrote it) records the split, so credit stays accurate as the lab grows. ## The decision, in short When a piece of work is done: 1. **Name the contribution.** Solver, analysis, reconstruction, theory, method, measurement, negative, tool, or exposition. 2. **Decide if it publishes.** Publish on rigor, not on excitement. A sound dead end publishes. A number with no re-runnable config waits. 3. **Open a PR at a tier.** Technical report if it is written up but not yet independently checked; reviewed finding once a second reader signs off. 4. **Place it.** Solvers and their kin under the experiments; structural results under why it is hard; techniques under building a solver; cross-solver measurements with the benchmarks. 5. **Chart it only if it is a solver.** That is the whole standard. It is deliberately light, because the point is to raise the floor on every page without slowing the notebook down. ## Related - [The lab](https://eternity2.dev/research/lab) — The wiki's open notebook: structural findings and named search experiments, each credited to the researcher who ran it and reproducible from source. One corner of the community's wider research. - [Experiments](https://eternity2.dev/research/lab/experiments) — The lab's named search experiments, one section per researcher. Each is a real run against Eternity II with its idea, its best board, and the questions it left open. Raphaël Anjou's notebook is here in full; the notebook is open to anyone else's. - [Contribute your research](https://eternity2.dev/research/contribute) — This wiki is the community's research home, and there is room in it for your work. Three ways to get it published, from a mailing-list post to a pull request, plus the small set of house rules that keep every page trustworthy.