Product leader · Tacoma, WA

I build AI products that turn messy, high-stakes data into decisions people can trust.

25 years building consumer and enterprise products, now leading AI-native software in a regulated industry. This site is where I explore on my own time, outside the day job. Every build here is a working system I made myself, shown with its evidence, its cost, and what it gets wrong.

Start with three builds Say hello

Raw data, before anyone makes sense of it.

25 yearsbuilding consumer and enterprise products

18working AI builds, each with its data, cost and failure cases

100,000+public records read by a model across these builds, the largest runs for a few dollars each

11essays on how product work changes when AI does the building

Start here

Three builds, one for each question the work keeps asking.

One question runs through the work: can a person trust what the AI did with their data?

Legible

Make the shape of complex data visible, so a person can see the whole before trusting any part.

Seen in Sorting Sea, The Eye, The Engine and the drug shortage study

Grounded

Every answer traces back to a source a reviewer can open and check.

Seen in Said vs. Filed and Can I Build Here?

Accountable

AI acts inside real systems, and a named person holds the decision.

Seen in Second Opinion, How Many Steps Does a Feature Need?, The Handoff Line and The Governor

All builds

Products I designed and built myself, on public data or tests anyone can rerun.

01Each one starts with the problem, sized with numbers anyone can check.

02It shows the math on what it’s worth, to whom, and whose work it changes.

03It says what it gets wrong, and what it hasn’t proven yet.

Writing

Short essays on how I build and what it taught me.

October 2026 · Interactive essayA second shift for every PMHow a product team can use the new always-on personal agents: jobs instead of vague assistants, one shared team memory, autonomy earned in steps, and what stays with people. October 2026 · Interactive essayWhen the pipeline collapsesFrom six AI-assisted handoffs to one ask and one acceptance: what has to be true first, what stays human, and what happens to the middle of a software company. October 2026 · Interactive essayWhen intelligence is abundantWhat stays scarce as models get much smarter: knowing what you want, ground truth, access, judgment and economics, and a 90-day plan to prepare. September 2026 · Interactive essayLearning at the speed of releaseKeeping up with AI without drowning in it: a filtered intake, a weekend lab, and a monthly habit of harvesting what experiments teach. September 2026 · Interactive essayTaste is the moatDesigning for jobs, not features, making every addition name a subtraction, and encoding taste into the skills agents build with. September 2026 · Interactive essayThe customer in the loopAmplifying real customer signal into panels that rank ideas and test designs while they’re built, and where real people stay essential.

About

I’m Nathan Rozendaal. I’ve spent 25 years building consumer and enterprise products, and today I lead product for AI-native software in a regulated industry, where every record can be audited and every change has to be explained. That habit of proving things follows me into every build here.

My habit is to build the working thing first, then write down what it taught me. This site is that record, kept on my own time and separate from my day job. Based in Tacoma, Washington.

Say hello

Send me a note

A question about one of the builds, a correction, or an idea worth trying. It comes straight to my inbox, and I read every one.