Contributor guide: how to publish research here

This is an open lab. Reading the research is free; so is writing it. This guide walks the whole lifecycle — from a rough idea to a published, peer-reviewed piece that others can build on.

You don't need a PhD, an affiliation, or a finished result. A careful negative result, a clean replication, or a sharp question are all worth publishing.

The lifecycle

1. Prepare — decide what you're actually claiming and gather your evidence.
2. Draft — write it up in Markdown, privately, in your own editor here.
3. Submit — send it for review when it's ready for other eyes.
4. Review — a maintainer reads it and either publishes it, asks for changes, or (rarely) declines it with a reason.
5. Revise — if changes are requested, edit and resubmit. There's no limit.
6. Publish & iterate — once live, it's open for the community to read, discuss, and reproduce. Living papers can keep improving.

Start a draft any time from Write research. Drafts are private to you until you submit them.

1. Prepare

Before you write a word, answer three questions in one sentence each:

• What is the claim? "X improves Y under condition Z" — not "some thoughts on X."
• What is the evidence? The experiment, the data, the argument, or the proof.
• Why should anyone trust it? What makes this reproducible or checkable.

If you can't fill in all three, that's fine — you may have a question rather than a result. Publish it as a question, and say so.

2. Draft

Write in Markdown. A good research note usually has this spine (see the research-note template for a copy-paste starting point):

• Summary — the claim and the takeaway, up top, in plain language.
• Background — what's known, and the gap you're addressing.
• Method — exactly what you did, in enough detail to repeat.
• Results — what happened, including what didn't.
• Limitations & next steps — where this could be wrong, and what's next.

Write for a sharp reader who is not in your subfield. Define jargon once. Prefer concrete numbers to adjectives.

The rigor checklist

Hold your draft to this before you submit. It's the same bar we hold our own work to.

• Claim is falsifiable. A reader can imagine evidence that would disprove it.
• Method is reproducible. Environment, data, and seeds are pinned — see reproducible experiments.
• Evidence matches the claim. You're not generalizing past what you measured.
• Baselines are fair. You compared against something a skeptic would accept.
• Negative space is stated. Failure modes, caveats, and compute budget are named, not hidden.
• Prior work is credited. You've linked what you built on.
• Numbers are checkable. Tables, configs, and (where possible) a link to code or artifacts.

3. Submit

When the checklist passes, hit Submit for review. Your piece moves to In review and locks for editing while a maintainer reads it. You'll get an email when there's a decision.

Small and finished beats big and vague. It's better to publish a tight result now and extend it later than to sit on a sprawling draft.

4–5. Review and revise

A reviewer will do one of three things:

• Publish — it's live on /research and open for discussion.
• Request changes — you'll see a note explaining what to address. Edit and resubmit; there's no penalty and no limit.
• Decline — rare, and always with a reason. You can revise and try again.

Reviews focus on rigor and clarity, not agreement. We publish things we disagree with when the work is sound.

6. After you publish

Published research is a living thing here. Respond to comments, fold in reproductions, and update it as you learn more. The goal isn't a frozen PDF — it's a result the next person can stand on.

Ready? Start a draft →