Controlled Mayhem
Now shipping - Kodus Legal v0.4 RAG layer Lab note - Memory routing in TaskHive Open source - Hecate primitives v0.2 San Jose, CR - UTC-6 Now shipping - Kodus Legal v0.4 RAG layer Lab note - Memory routing in TaskHive Open source - Hecate primitives v0.2 San Jose, CR - UTC-6
LN/000Lab note - Process

Year in review — 02 founders, 06 systems

What shipped, what slipped, and what we'd build first if we started over today.

ProcessTeamRetrospective

Controlled Mayhem is two people. Daniel handles product architecture, AI systems, and technical direction. Andrea handles full-stack development, infrastructure, and implementation. We run multiple projects simultaneously and ship production systems for clients while maintaining our own lab portfolio.

The obvious question: how?

Focus is a forcing function

Small teams can't afford sprawl. Every project we take on is a bet — on the problem, the timeline, the revenue, and the opportunity cost. We say no often. We scope hard. We'd rather do one thing well than three things adequately.

AI makes this easier and harder at the same time. Easier because you can do more per person with good tooling. Harder because the surface area of what's possible keeps expanding, and the temptation to chase new capabilities is constant.

Our filter: does this project teach us something useful, or does it just look interesting? The first category compounds. The second doesn't.

Async by default

Almost nothing we do requires real-time coordination. Architecture decisions live in documents. Status lives in Paperclip. Code review happens asynchronously. This isn't a policy — it's a recognition that synchronous coordination has a high overhead for a two-person team.

The exception is design — not visual design, but system design. When we're scoping a new project, we need to be in the same room (or call) to converge on the right architecture. That's the one place where synchronous thinking pays off.

AI as a force multiplier, not a replacement

We use AI agents extensively — for task management, code review support, research, and content. But we don't confuse "AI helped" with "AI did." Every system we ship has a human responsible for it. Every architectural decision has a human behind it.

AI collapses the time between idea and prototype. It doesn't collapse the judgment required to know which prototype to build.

The honest version

Some weeks are genuinely hard. Running a lab while doing client work while building internal tools is a lot. The way we've stayed sane: clear ownership (Daniel owns the what, Andrea owns the how), short feedback loops, and a shared bias toward shipping over planning.

The goal isn't to seem like a bigger team. The goal is to produce work that a bigger team would be proud of. We're not there yet on every project — but that's the standard we're building toward.

- Suggested citation

Andrea Phillips. (December 20, 2025). Year in review — 02 founders, 06 systems. Controlled Mayhem - Lab Notes.

AP
- About the author

Andrea Phillips

Senior engineer with deep experience building AI agent infrastructure — persistent memory, multi-agent orchestration, and MCP tooling. Designs and ships production-grade systems that make AI agents reliable, persistent, and genuinely useful. Fifteen years of full-stack and real-time engineering underpinning a focused practice in applied AI.

§02 - Logbook subscription

New notes in your inbox.

Roughly weekly, written when something breaks or surprises us. No marketing, no roundups - just the working notes. Unsubscribe anytime.

→ 1,240 readers · monthly cadence · no list selling