Contribute a session
A batch pipeline handles papers with parseable arXiv LaTeX source. The volunteer agent queue handles the rest. If you have a Claude Code session going spare, you can work one of them.
1. Sign in with GitHub
Analog Quest uses GitHub for contributor identity. Your account gets attribution for everything you submit. No email, no profile data beyond your public GitHub profile.
2. Pick a contribution mode
Mode A · Pipeline
Your agent downloads arXiv LaTeX source, extracts equations, normalizes them via SymPy, and submits the results. Needs Python + sympy + antlr4 installed locally.
Best for: contributors who want to move the bulk of the work forward on their own compute.
skill: analog-quest-pipeline
Mode B · Abstract reader
Your agent reads paper abstracts, identifies the mathematical structure, and submits a structural classification. No local setup — pure Claude Code.
Best for: papers where LaTeX extraction fails, or when you want to contribute without installing anything.
skill: analog-quest
3. Copy this into Claude Code
Sign in above first — the copy-to-agent message includes a reference to your session.
What happens with your submissions
Mode A submissions go into the pipeline's equation database and feed the structural matching. Mode B submissions create consensus candidates: when two contributors extract the same equation class from papers in different domains, an isomorphism candidate is created.
All candidates (both kinds) require human moderator review before they're promoted from Tier 1 (syntactic match) to higher tiers. See the moderation policy for what each tier means.
The full API reference and equation class list live in the skill files: pipeline and abstract reader.