PROJECT · active

Naulus

A public evidence map for fitness and nutrition claims, where nothing publishes with an unverified citation.

Next.js · Supabase · extraction pipeline · OpenAlex · Crossref

NaulusEvidence, mapped

Fitness & Nutrition · The Evidence, Mapped

Claims, mapped to their evidence.

Getting lean isn’t a mystery. It’s a settled question buried under noise.

Every citation verified · nothing publishes unverified
Why it has to be right

Health advice with invented sources is the internet's default. Naulus verifies every citation before it publishes.

What it is

Naulus maps popular fitness and nutrition claims to their actual evidence. Pick a claim, see the verdict, the studies behind it, how strong they really are, and what the practical takeaway is. It exists because I needed it myself: I was rebuilding my own health and kept finding that the loudest advice had the weakest sourcing.

The evidence problem

Fitness content is an evidence minefield. Creators cite studies that don't say what they claim, mechanisms get presented as outcomes, and a 12-person trial becomes a headline. The failure mode is identical to AI hallucination: it sounds right and isn't. So I built the same defense I use in my research pipelines.

The claims pipeline

The claims pipeline extracts assertions from sources, grades the supporting evidence, and publishes claim pages with full provenance. The site is information-first, with no login. There's no program for sale and no affiliate links, just claims, evidence, and a methodology page that shows how the sausage is made.

Keeping the numbers honest

Every citation on every claim page passes through the same machine-verification gate as my research pipeline: parsed, validated against OpenAlex and Crossref, quarantined on failure, timestamped on success. A human (me) approves every claim node before it goes live. The publishing rule is absolute: no verified citation, no publication. The other deterministic core: any computed number on the site comes from plain code, never from a model. I don't let the model author the numbers.

What broke and what I learned

The first version of this project was a coaching app. I built it before asking who it served. Reframing it as an evidence resource made it both more useful to other people and more honest about where I am: mid-journey, tracking my own training and meals with the same rigor I expect from the studies.

Status

Active, rebuilding in public. New claim nodes are added in batches with a public changelog.