The short version
I've spent 30 years in IT — infrastructure, security, automation. I know what needs to exist and why. I can spec it, architect it, and ship it. What I can't do is write production code from scratch the way a software engineer can. And I have enormous respect for that craft.
I also have enormous respect for product managers. I borrow from their playbook — user needs, prioritization, scope — but I haven't earned that title either.
Here's what I think we all need to get comfortable with: AI is creating a new role that sits between disciplines. It doesn't replace developers. It doesn't replace PMs. It lets domain experts ship in ways that weren't possible two years ago.
That's not a threat to anyone's craft. It's a new seat at the table.
The coders, the PMs, the builders — we're all going to need each other more, not less. The work just got bigger.
The receipts
DevPlan MCP Server
AI coding assistants lose context between sessions, skip steps, and produce inconsistent code. You end up debugging for hours, repeating the same mistakes, with no record of what was built or why.
DevPlan fixes this. It's an MCP server that brings structured planning to AI coding tools — Claude Code, Cursor, Aider, Cline, and Windsurf. It interviews you about your project, generates a validated development plan, then hands it off to AI to execute mechanically while a separate verifier actively tries to break the result.
Issues found during verification become lessons that automatically improve future projects. It gets smarter every time you use it. Open source, MIT licensed, runs on Cloudflare Workers.
amp-rs
AI agents forget everything between sessions. Every conversation starts from zero — no memory of what worked, what failed, or where you left off. Existing solutions either dump the entire context (expensive, noisy) or lose it entirely.
amp-rs is the Rust reference implementation of Agent Memory Protocol. It gives AI agents structured, persistent memory through three primitives: checkpoints (where you left off), lessons (what you learned), and memory (facts worth keeping). Local-first, fast, and built to be embedded anywhere.
Built in Rust for speed and reliability. Runs locally — no cloud, no subscription. Powers Nellie, my personal code memory server. Open source, MIT licensed.
More projects
MESH Protocol
→Memory Exchange & Sharing Hub — a secure federation protocol for AMP nodes. Share agent memory across teams and organizations without giving up control of your data.
Praxis
A system for turning dense documentation into expert systems that learn. Three layers: foundation docs, institutional knowledge, and an AMP-powered refinement loop that compounds with every correction. Working demos in CMMC compliance, mortgage underwriting, and Bay Area real estate — with any document-heavy vertical as a candidate.
Frank is a Bay Area realtor. A buyer calls — wants to see a property in Oakland in an hour. Frank calls the office: "dump the disclosures for 4th Street into Praxis." By the time he pulls up, he's got rent control status, soft-story retrofit requirements, lead paint obligations, and every city-specific disclosure at his fingertips. He asks Praxis questions on his phone and walks in like he's been studying the property for a week.
Latest thinking
Are We Solving Context Wrong?
The AI industry keeps pushing bigger context windows. But context reset isn't a capacity problem — it's a signal problem.
Let's talk
I'm looking for my next role — somewhere that values this kind of hybrid builder. If that resonates, reach out.