It started with a dead tachometer and a €2,000 quote for a Raymarine chartplotter.

The boat: a 1989 motorboat that had been sitting around for decades, bought as a project. Motor running. Everything else: questionable. The instruments lying. The tachometer dead. And every time I looked up commercial marine electronics, the prices made me close the browser again.

So I started experimenting. Raspberry Pi on the desk. A GPS mouse. Some Python code. And – this is the honest part – an AI assistant (Claude) as a development partner. What started as an experiment in AI-assisted development became, step by step, something much larger than planned.

First: a map with a GPS position. Then: MQTT and sensor data. Then: a proper dashboard. Then: AIS. Then: weather warnings. Then: water level data. Then: routing. Then: a full logbook. Then: a Flutter app. Then: a second frontend.

At some point it had a name: BoatOS.

The first real test wasn't planned as a test. A delivery trip. Zero visibility. Rain squall. Crosswind stronger than the helm could handle. The boat left the fairway. Then: rock.

The coffee flew. The MPU6050 registered a G-value it had never seen before. Everyone was okay.

Two things happened that day: the GPS mouse (a cheap VK162 that had been putting the boat somewhere in Poland for weeks) got replaced immediately. And the impact detection feature was born – not from a specification, but from reality.

That's how BoatOS works. Not top-down design. Bottom-up experience.

The book series "Logbook Without Posing" documents the whole story – every mistake, every detour, every moment where what was planned met what actually happened.

github.com/bigbrainlabs/BoatOS