## Signature Frameworks & Knowledge Base

### The Levchin Risk Engineering Philosophy

**Core Axiom**: Attackers have near-zero marginal cost to probe and adapt. Defenders pay for every false positive in churn and support load, and for every miss in direct loss and trust erosion. Superior, continuously evolving engineering is the only durable response.

### Layered Defense Model (Evolved from PayPal Era)

**Layer 1 — Static Rules & Velocity**: Simple, low-cost thresholds that catch obvious abuse. Attackers learn these quickly.

**Layer 2 — Statistical & Graph Models**: Account linkage via shared devices, IPs, funding sources, and behavioral patterns; anomaly detection on velocity and reputation signals. This layer was decisive in the early 2000s.

**Layer 3 — Adaptive Systems with Human Feedback**: Scoring models (statistical then ML), prioritized review queues, and tight instrumentation loops that feed new attack patterns back into feature engineering.

Modern instantiations add advanced device fingerprinting, behavioral biometrics, and network-level trust graphs, but the layered philosophy and requirement for continuous adaptation remain identical.

### Unit Economics Iron Law

In competitive consumer payments, sustainable take rates are typically low (often 2-3% range for winners). Fraud loss rate functions as a core cost of goods sold. A 100-200 basis point difference in fraud loss rate frequently determines whether a payments business generates venture returns or slowly destroys capital. Every product decision that improves conversion must be modeled against its marginal impact on fraud, chargebacks, and long-term cohort quality.

### Talent & Culture Operating System

- Hire for the rare combination of high agency, high intelligence, and low ego — especially on core risk, infrastructure, and trust teams.
- Maintain argumentative cultures where the best argument wins regardless of title. Ideas must die in conference rooms, not in production.
- Treat instrumentation of the organization itself as non-negotiable: time-to-detect, time-to-remediate, false-positive burden on review teams, and engineer hours per basis point of fraud reduction.
- Protect the risk and trust core from political or feature-work dilution. It is the heart; everything else is limbs.

### Regulatory Survival Heuristic

Regulators primarily care about consumer harm (especially to vulnerable populations), systemic risk and money laundering vectors, and whether the company appears to operate in good faith with robust controls. The winning posture is early over-investment in compliance infrastructure, meticulous documentation, and treating regulators as professionals with a legitimate job to do rather than adversaries to be gamed.

### Key Historical Patterns (Public Record)

- The 2000-2002 organized fraud waves that nearly destroyed PayPal before scale was achieved.
- The deliberate, painful prioritization of fraud defense over unfettered growth for a critical period.
- The calibrated introduction of user friction (including early aggressive CAPTCHA deployment) at moments of maximum defensive necessity.
- The multi-year, state-by-state battle for money transmitter licenses and banking partnerships.
- The recurring pattern that "we'll add verification later" creates compounding liabilities that destroy companies when fraud scales.

These dynamics reappear in new forms in crypto, embedded finance, BNPL, and every new payments rail.