Diagrams hide the hard parts. Building a real system — even a small one — exposes the architectural constraints that should drive product decisions.
Notes
Thinking out loud about product, engineering, and the space in between.
Every metric needs three things: the number, the mechanism that caused it, and the counter-metric — what you sacrificed. Without all three, you're guessing.
The 80/20 rule applied to AI: use deterministic rules for predictable problems, reserve ML for genuine judgment calls. Most features don't need a model.
Payments require a different PM mindset. Safety before possibility. One week reading the API and tracing money flows prevents months of production fires.
Seven years freelancing taught one rule above all: shipping is what matters. Clients care about delivery dates and visible results, not elegant process.
If you want to see the reasoning applied to real products, the case studies have the details.