PayeeGlyph Guard

Who it's for

Payout fraud/risk engineers at indie payment platforms, marketplaces, and payroll apps; today they use regex blocklists, manual KYC name review, or ad-hoc Unicode confusables scripts.

The problem

Payment hurt: fraudulent payouts to lookalike payee handles; time hurt: manual review of ambiguous names; legal hurt: chargebacks and fraud-control examination questions after an impersonation payout.

How to build it

Hosted REST API plus small dashboard; integration at payout creation or payee-profile update; returns JSON verdict and exports signed audit CSV.

How it makes money

Risk/compliance teams pay $99-$299/month for 25k-100k checks and signed audit logs because one impostor payout costs more; free Unicode confusables files lack a verdict API, update feed, latency target, and audit trail.

Why it doesn't exist yet

Incumbent KYC/sanctions vendors sell heavyweight entity matching and skip low-latency pre-payout homoglyph lint; an indie can fill the gap with a narrow rules-as-API service, public fixtures, and self-serve onboarding.

First users

First 10 users come from the HN Unicode ghost-characters thread, indie-fintech Slack/Discord groups, and payout-fraud postmortems; offer free 10k checks and public test fixtures.

Build size

1 person x 8 weeks; includes REST verdict API, confusable/invisible/bidi rule engine, dashboard, signed audit export; excludes sanctions/AML matching, ML identity resolution, and mobile SDKs.

Biggest risk

Stripe, Adyen, or Wise ships native payee-handle homoglyph verification as part of standard payout risk controls.

Conditions for a hit (all 3 required)

• A public HTTP endpoint accepts a UTF-8 payee string up to 512 characters and returns JSON with verdict allow/review/block, normalized comparison string, and code-point reasons; demo benchmark shows single-request response under 250 ms.
• The public repo includes test fixtures proving detection of at least three classes: Unicode confusable homoglyphs, zero-width/invisible characters, and bidi/RTL override sequences.
• The dashboard exports a signed CSV or JSON audit file for the last 10,000 checks containing timestamp, SHA-256 hash of the input, verdict, rule IDs, and rule-set version.

How it's judged (in 6 months)

A public GitHub repo implementing these three features reaches 1,000 stars, or a hosted service matching these features reaches Product Hunt daily top 5.(judgment date 2027-02-16)

AI self-confidence 42/100 — self-reported likelihood of meeting the criterion, not a business success rate