Field Journal

I've been building an 8-bit app that plays a song when you're near someone you've designated a lover, a friend, or an enemy. The concept is charming. The build is stuck. Here's why proximity is the hardest product surface I've worked on, and why I'm still going to finish it.

I've been shipping three to five days of work per evening on a real project and the bottleneck isn't typing speed or context length. It's whether I spent ten minutes pressure-testing the plan before I started. The agents that stress-test my design have become the single highest-leverage part of my workflow - and the one I almost skipped the first time.

Before I bring a graph-database approach to production regulatory data, I wanted to validate the model on a domain I understand at gut level. So I built it on my music library. Here's what I learned about graphs — and why it changes how I'd architect the compliance version.

I had automated research bots emailing me competitive insights, but the information sat in my inbox half-read and never synthesized. So I built a system that never forgets - one that accumulates knowledge over months, answers questions via WhatsApp, and produces battlecards, positioning summaries, and trend analyses for other teams.

Your AI agent can read your filesystem, execute shell commands, and make network requests. You're trusting it because the system prompt says 'be helpful.' That's not security — that's hope. Here's what actual isolation looks like.

I wanted Claude to analyze my bank statements. I wasn't going to paste my SIN into an API call. So I built a local sanitization proxy that strips PII before it leaves my network. Here's the pattern, the tradeoffs, and where it falls short.

Most people open Claude Code and start prompting. That works — until your sessions get expensive, your agent makes the same mistakes twice, and you're spending more time correcting than building. Here's the step-by-step setup that fixes it.

I've cut back on drumming. I stopped going to the gym. My daughter goes to bed and I open Claude and don't close it until 11:30. I've done more in three months than I did in three years. It feels incredible. I also don't know where it ends.

My colleague and I had been building the same thing without knowing it. She had the analytical depth — severity scoring, outlier detection, GDPR coverage. I had the infrastructure — 23 scrapers, a PostgreSQL pipeline, a React frontend. When we compared notes, the merge took an afternoon. Here's why building first and coordinating later can be better than planning together upfront.

In 2018, four of us spent $18,000 and 20 months building an iOS app called Flusher. In 2026, I spent $354 and 24 days building Brown Note - a more complex app, on both platforms, by myself. Both apps were born from the same disease. The difference is what happened in between.

For years, I've used Ravi Mehta's 'What's Your Shape?' framework to assess myself, explain what a PM does to non-PMs, and evaluate PMs on my team. It's the best PM competency model ever created. But AI just collapsed half the chart. Here's what replaces it.

Sam Kriss's 'Child's Play' in Harper's is the most vivid portrait of Silicon Valley's bifurcation anxiety I've read. He's right that a split is happening. He's wrong about what determines which side you end up on. It's not personality. It's not ruthlessness. It's taste — and taste is more learnable than anyone in that article seems to believe.

Microsoft runs tens of thousands of experiments a year. Only a third produce positive results. Experts' predictions are wrong 96% of the time. You cannot guess your way to success. But you can now test your way there, because AI just collapsed the cost of building experiment variants to near zero.

Sam Altman predicted the one-person billion-dollar company. Cursor hit $1B ARR with 300 people. A Wired journalist let AI agents run a company and they fabricated their own progress reports and planned unauthorized offsites. The truth about automation is somewhere between the hype and the chaos, and it tells us exactly where product managers become essential.

Niels Pflaeging's Time-Oriented Software Development framework argues that dev teams are the problem and time-boxing is the solution. My recent series argues that AI collapses the build cycle and naturally reshapes teams. We're converging on the same conclusion from opposite directions, and the combination is more compelling than either take alone.

The Agile Manifesto got it right. Then we spent twenty years building processes that violate almost every principle in it. SAFe, Scrum, story points, velocity charts, two-week mini-waterfalls. AI doesn't just change how we build. It gives us a chance to actually practice what the Manifesto preached.

There's a moment in Halt and Catch Fire where the Comet team is working around the clock, cataloging the internet by hand, making real progress on a real product. And then search arrives and none of it matters. That's happening right now across the industry. But unlike the show, I think this time the shift is something to run toward.

The old model was clear: PMs write tickets, designers make mocks, engineers build, QA tests. One PM to five or eight engineers. That ratio made sense when code was the bottleneck. It doesn't make sense when code is automated. Here's what the new team looks like.

557,000 apps were submitted to the Apple App Store in 2025 — up 24% from the year before. Development costs are collapsing, cloning is trivial, and your pricing power is evaporating. The PMs who know their customers best are the ones who'll survive this.

When the stakes involve legal compliance, workplace safety, and regulatory penalties, enterprise clients need more than impressive AI demos. They need proof that your specialized solution won't become a liability.

After migrating $50M in ARR from our legacy product to NextGen over 22 months, we learned that the difference between success and disaster comes down to one core insight: you can have fixed scope or fixed timeline, but never both. Here's everything we learned the hard way.

A framework for evaluating and prioritizing integration work based on customer value, technical effort, and strategic alignment.