Building HatchClaw

R&D Project — 2025-2026

HatchClaw is one of several experiments I’m running to explore what AI-native product development looks like.

After 20 years in product leadership—8 as an engineer, 10 as a PM—I’m spending 2025-2026 testing a hypothesis: what happens when AI agents don’t just assist product work, they execute it?

This page documents what I learn. Honest, including failures.

The experiment

HatchClaw is a multi-agent system for building software. The idea: what if your product team included AI agents that could actually do the work—not just suggest, but execute?

An AI CMO that handles positioning and messaging
An AI CTO that makes technical decisions
AI product managers that run discovery and write specs
AI engineers that ship code

The core question: what does the human do in this system?

So far, the answer seems to be: judgment, taste, and system design. The human becomes an architect, not a task manager.

Why this matters

I’ve watched product work evolve over two decades. The current shift—where AI dissolves the boundary between product and engineering—is the biggest change I’ve seen.

Traditional product management assumes humans do the execution: write specs, manage roadmaps, prioritize backlogs. But what if AI handles that? What does the PM become?

HatchClaw is my attempt to find out—by building the future I’m trying to understand.

Building in public

I’m sharing everything:

Monthly metrics — honest numbers on what’s working
Technical decisions — architecture choices, pivots, mistakes
Product learnings — what users want vs. what I thought they’d want
Operational challenges — running a “company” where most of the team is AI

Latest snapshot

HATCHCLAW SNAPSHOT (Month 1)

Users: Building
Biggest win: Getting the multi-agent architecture working
Biggest surprise: How hard it is to coordinate AI agents
What’s next: First user-facing features

Part of a bigger exploration

HatchClaw is one of several R&D experiments:

→ HatchClaw — Multi-agent system for building software
→ Concierge of AI — AI-assisted research and synthesis
→ Vockify — Voice-first AI interactions

These aren’t side projects. They’re deliberate experiments in what stays, what goes, and what changes as AI becomes the executor.

The ethics of building in public

I believe in transparency, but I’m thoughtful about:

What metrics I share (user numbers yes, revenue TBD)
User privacy (anonymized learnings only)
Competitive considerations (general strategies, not specific tactics)

Follow along

→ Visit HatchClaw
→ Read The Boundary — weekly updates on what I learn
→ Follow on LinkedIn — shorter updates and insights

This page is updated monthly with new snapshots. Last updated: February 2026.