Skip to content

Project Standards

I Love PETs is a curated field guide. Its value comes from judgment that readers can inspect, not from collecting every link.

Quality Bar

Every decision-support page should help a reader answer:

What should I believe, why should I believe it, what might have changed, and what should I verify before building?

Good pages are practical, dated, scoped, and honest about failure. A page is not ready if it only explains what a PET is.

Required Page Qualities

Quality Standard
Decision support The page says when to use, when not to use, and what would change the recommendation.
Threat model The page names adversaries, protected assets, assumptions, and outputs.
Failure modes The page explains how the PET can fail or be misused.
Evidence Claims are sourced, measured, labeled as expert judgment, or placed in the claim register.
Specificity Examples name actors, data, outputs, and operational constraints.
Anti-hype The page does not imply a PET solves governance, security, utility, or output leakage by itself.

Minimum Useful Page

Every page should contain at least one of:

  • decision table;
  • worked example;
  • failure mode;
  • checklist;
  • concrete research problem;
  • tradeoff matrix.

Writing Rules

Prefer:

  • "Use this when..."
  • "Avoid this when..."
  • "This breaks when..."
  • "This does not protect against..."
  • "A good first experiment would be..."
  • "The hidden cost is..."

Avoid:

  • generic textbook prose;
  • unqualified "privacy-preserving" claims;
  • paper dumps;
  • tool lists without fit, threat model, or limitations;
  • empty sections waiting for future content.

Review Gate

Before opening a PR, check:

  • mkdocs build succeeds.
  • Navigation still reaches the changed page.
  • Internal links point to existing pages and anchors.
  • No empty headings remain.
  • The page contains at least one concrete decision aid.
  • Any deployment or tool claim is sourced or explicitly labeled as uncertain.

What Counts As Progress

Improving one important page is better than adding five shallow pages. A useful PR makes a reader more capable of choosing, rejecting, evaluating, or debugging a PET.