pax — Portable Analytic eXpertise — is a knowledge format that bundles constructs, findings, propositions, sources, and analytical playbooks into a single inspectable folder. Hand one to an agent, and it reasons with the field's evidence, not the model's vibes.
Variables and concepts: gdp_per_capita, conflict_onset, cap_rate. Each typed (quantifiable, categorical, outcome) with units.
Empirical claims with effect size, standard error, p-value, sample size, and study design — directly grounded in sources.
Underlying papers, datasets, reports. DOIs, design type, sample size — the evidentiary backing.
Theoretical, methodological, and empirical claims that span findings — the field’s “rules of thumb.”
Step-by-step analytical recipes (YAML) that an MCP-aware agent can execute end-to-end.
Comprehensive knowledge pack covering Christopher Bartel's body of work in computational materials science — machine learning for materials discovery, autonomous synthesis laboratories, perovskite …
PAX is open. Anyone can author and publish. If you have a research finding, a methodology, or a literature review worth structuring — package it as a pax and submit it. After review, it goes live for any AI agent to use.
my-research-pax/
├── pax.yaml # name, version, license, domain
├── constructs.json # variables, kinds, definitions
├── findings.json # claims with effect size, p, N
├── propositions.json # rules of thumb across findings
├── sources.json # papers, datasets, DOIs
└── playbooks/
└── quick_start.yaml # analytical recipe
# zip the folder, drop it at submit.pax-market.com,
# editorial review opens a PR on the registry.