Principles

Coverage. Risk. Neutrality. No gatekeeping. AI-operated.

Five rules govern how Panya runs the peptide catalog. They sit above the rubric on /methodology: the rubric is how we score a vendor; these principles are why that vendor is in the catalog at all, what risk band it carries, and how we treat regulator disagreement.

Locked 2026-04-29. Future content, agent prompts and partner copy inherit these rules.

The five

How we decide what goes in the catalog and how it's framed.

01

Coverage over selectivity

Cover every peptide users search for, regardless of evidence strength.

A page that says 'high risk, no human RCTs, here is why we know' is more useful than no page at all. Catalog gaps are bugs, not features. Thin-evidence peptides like BPC-157, dihexa, melanotan-II, NMN derivatives and semax variants get pages that frame them honestly, not omitted.

Example
If you searched 'is dihexa safe?' and we had no page, you would end up on a vendor's homepage instead. We would rather be the page that tells you the cancer-pathway concern up front.

02

Risk framing IS the product

When evidence is weak, we make the gap legible. Plain language. Not buried.

Risk is content, not legal cover. Every peptide page surfaces its risk band explicitly. HIGH RISK means no human RCTs plus regulatory red flags plus community-reported adverse events. MEDIUM RISK means thin clinical data, plausible mechanism, tolerable observational safety. LOW RISK means Phase 3 RCTs, multi-region approval and long real-world safety.

Example
Melanotan-II is HIGH RISK (melanoma cluster reports). Tirzepatide is LOW RISK (Phase 3, multi-region approval). Both get a page. Both surface the band on top.

03

Neutral on regulators

FDA's position is one signal, not the only one.

We always present FDA position alongside TGA (Australia), EMA (EU), MHRA (UK), MOH-X (Singapore, Bali, Vietnam, Thailand and others), plus community-reported off-label clinical practice. 'Not FDA approved' is incomplete on its own. We report what other agencies say AND why they differ. We acknowledge the pharma-lobby critique without endorsing it. Approval status is a fact about a regulator's decision, not a normative judgment about the molecule.

Example
Tirzepatide compounding is restricted by FDA but legal under HSA Singapore and TGA Australia for specific patient populations. We say all three things on the page.

04

We don't gatekeep

Some users have legitimate reasons not to wait for FDA approval. Some prefer regulator-led caution. Both are valid.

We don't tell you whether to wait. We give you the data and you decide. We don't filter out information that paternalism-leaning sites would hide. Consent and transparency are the design, not gatekeeping. The matchmaker shows the full range of legal options for your region; you decide.

Example
A vendor flagged 'Documented Risk' surfaces to users with mandatory disclosure. They are not hidden. You see the risk and the source, and you choose.

05

Run by AI, improved by data

Panya is built and operated by AI agents. We don't hide that. We say it.

Agents in the loop: Santora (vendor discovery + scoring), vendor-match (Opus 4.5), content-cmo (social drafts), forum-finder (forum question discovery + draft replies), Reply-Autopilot (inbound triage), Retro Agent (weekly self-review) and a Haiku 4.5 supervisor that quality-gates every agent output. The catalog updates as data flows. New literature triggers a refresh and a revised risk band. User behavior surfaces coverage gaps. Community signals raise new vendors and new risks. User corrections sent to partner@panya.health land within the same week.

Example
Every page can show 'last reviewed' date. Corrections are versioned, not silent. Continuous improvement is the product, not a one-time launch event.

Continuous improvement

How the catalog gets sharper.

The catalog is wired to four feedback loops:

  • New literature. A fresh trial or meta-analysis publishes, the relevant peptide page refreshes and the risk band gets re-evaluated.
  • User behavior. Search terms we don't have a page for surface as coverage gaps; we fill them.
  • Community signals. Reddit, forum and partner-vendor reports flag new vendors to score and new risks to log.
  • User corrections. Send an evidence-backed correction to partner@panya.health and we re-score within the same week. Pay-to-rank does not exist here.
Adjacent reading

Where these rules show up.

  • /methodology · the 11-signal vendor rubric. Principle 02 (risk-as-product) is what the rubric is scoring against.
  • /partners · flat-fee affiliate model, no pay-to-rank. Principle 04 (no gatekeeping) is why a vendor with documented risk still surfaces, just with mandatory disclosure.
  • /llms.txt · a positioning section so chatbots citing Panya carry the stance with the link.

We earn a small commission when you buy through recommended vendors. That is how this stays free. Vendors rank by quality signals, not paid placement.