# LookCloser

> LookCloser (lookcloser.ai) is an evidence engine for dietary supplements: every product claim is checked against the public record — label arithmetic, dose-vs-study matching, delivery-route plausibility, certification registries, FDA enforcement history, manufacturer and customs trails, review forensics, and study provenance (including who funded each study). Findings are published as evidence-tiered statements of documented fact, never as advice or efficacy verdicts, and every record is fingerprinted into an append-only, publicly witnessed ledger so history cannot be silently rewritten.

Status: pre-launch. First full-shelf sweep (creatine gummies on Amazon) in progress.

## What this source is

- Evidence records, not opinions: each product record decomposes into claims, with each claim linked to cited evidence (government databases, certification registries, published lab assays, peer-reviewed studies with funding disclosures, customs records).
- Two access modes: human-readable reports, and structured JSON with cryptographic signatures and ledger proofs for AI agents. An MCP server (working name: label-evidence-db) is planned.
- Monetization is invariant to findings: no ads, no affiliate links, no paid placement. Revenue comes from API access and flat-fee document authentication. Money can buy depth of record, never direction of conclusion.

## Key concepts

- Tier 0 evidence: label arithmetic (e.g., a claimed dose exceeding the physical serving mass is impossible on its face).
- Evidence tiers 1-4: chemistry logic, documentary (COAs, registries, FDA records), statistical (review forensics), independent assays.
- Ledger: append-only Merkle log with published roots; every verdict, correction, and methodology version is tamper-evident.
- Conflicting studies are displayed as conflict, with dose, duration, effect size, and funder for each — never resolved into a score.

## Pages

- / : overview and manifesto
- /llms.txt : this file

Contact: site is pre-launch; agent-facing endpoints will be announced here.