Notes on building AI tools that people actually use.
I'm an independent developer and founder in Boulder, Colorado. I build AI products end to end — a portfolio of consumer AI web tools under CLOOMS, and AI-first B2B software under NexaPioneer, running on a database of 70M+ integrated B2B records. I write here about what actually works when you ship AI tools: picking the right problem, wiring models into a real product, and holding quality once real users show up.
Recent Articles
Part 2 of a deep read of Google's Search Quality Rater Guidelines: the Needs Met scale, query intent types, why 'Do' queries change how tool pages are judged, and a working audit for AI tool sites.
A deep read of Google's Search Quality Rater Guidelines (Sept 11, 2025) for people who ship AI tools: the three-layer model, E-E-A-T with Trust at the center, and what actually gets a page rated Lowest.
My build filter: name one specific job in one sentence, or don't ship. How I choose what to build under CLOOMS — and what I kill before it starts.
What I learned building and running a set of single-purpose AI web tools under CLOOMS — on models as commodities, quality gates, and why the first ten seconds decide everything.
