Senior Software Engineer at Nexus Labs. Primary engineer for Maverick Protocol, a DeFi trading platform (80K peak DAU, ~3M users, $70B+ cumulative volume, 6 chains). Over 4 years I progressed from team contributor to sole owner of the V2 stack (98% of API, 73% of frontend), shipping 4,226 commits across 12 repos, 121 production releases, and 3 additional DeFi products. PhD Mathematics (Georgia Tech). Former department chair and teacher for 15 years. Head rowing coach, #1 nationally ranked masters women's program (2025).
Web3-Claude — An open-source agent framework for Claude Code. The foundation for a production swarm of 100+ specialized subagents that coordinate parallel work, enforce architecture decisions, and maintain code quality across a Web3 codebase. The system routinely compresses multi-week engineering efforts into a single day.
johnpphd.com — Writing about agentic engineering, AI agent patterns, and lessons from building production agent systems.
eslint-plugin-harness-guardrails — ESLint rules that catch the failure modes AI is fastest at producing: components with too many hooks, dispatchers with too many branches, positional data leaking across boundaries.
eslint-plugin-branded-types — ESLint rules enforcing branded type safety. Bans as casts, direct construction, and mutation of Zod branded types.
claude-code-skill-statusline — Show which skill is active in your Claude Code statusline, with session isolation for multi-session projects.
TipNote — A Notion-like collaborative document editor built with React, TipTap, and Firebase. Includes an MCP server for AI integration.
IsThisPersonAScam.com — AI-powered scam detection. Paste a LinkedIn profile or message, get a risk verdict in seconds.
- Failures Become Guardrails
- Real Programmers Don't Use AI
- Deterministic Scaffolding
- AI Is Not Logical
- Speed Is Not Velocity
- Build the Demo, Skip the Deck. Then Delete Half of What You Built.
- Five Theories of Agent Skills
- The Spec Is the Hard Part
- The Illusion of Control: Building a 100+ Agent Swarm in Web3 (Part 3) — HackerNoon
- You Are Part of the Harness
- Every Senior Developer Believes in DRY
- Your Agents Violate Your Principles
- The Illusion of Control
- The Evolution of AI Users
- You Think You're in Control?
- Ship Types, Not Markdown
- RL Know vs RL Do: Why AI Learned to Act
- SkillsBench: The Skills Advantage
- Why I Decided to Build 100+ AI Agents for Web3 — HackerNoon
- Lessons from Building a 100+ Agent Swarm in Web3 — HackerNoon (front page)
- 5 Rules for Agentic Engineering
- DRYP: Don't Repeat Your Prompt
- More Context Is Not Better
- Skills vs Agents