Research
is
where
I
start.
Building
is
how
I
learn
what
I
missed.
Twelve years designing. Eight in research. Now solo-building AI-native products where the design is what makes the AI honest.
AI for urban-code compliance. Reads a building project, cross-references it against municipal code, tells the architect if the design will pass, before city hall says no.
Pre-investment terrain viability. Multi-scenario reports for investors deciding whether to buy or develop a plot, investor language, minutes not weeks.
AI music-taste mapper for people whose Spotify algorithm has flattened them. Reads your library, reveals the shape of your taste.
AI that shows its work and asks when it doesn't know. Source tags over confidence theater. The product doesn't replace the expert, it gives them an auditable triage.
Qualitative findings as atomic, queryable units. Same logic from Sheets-by-hand to LLM agents. Method is what compounds, tools are throughput.
Tools change every year. Methods compound for decades. I bet on the practices I'll still be using in five.
What a project leaves on the table tells you more than what it ships. The messy middle is a senior move, not a junior risk.