Where I write up what’s actually happening when I try to run things — a dance community, a civic-accountability investigation, properties — AI-native. What’s working, what isn’t, what I haven’t figured out yet.
No takes, no thought-leadership — just the small “oh, this is now SUPER easy” and the equally important “this still doesn’t work.”
Projects#
- Running a dance community — attendance pipelines, per-tier outreach, the website I forgot I had.
- Keeping the city accountable — Public Records Act forensics on a 34-month-late public works project. Three LLMs in parallel, journalist outreach, the slow grind.
- Property management — one remote cabin, one local house. Two vendor stacks. AI doesn’t run the vendors; it runs me.
Redwood City’s Veterans Memorial Senior Center Phase 1 came in 34 months late at significant public cost. I filed a Public Records Act request, got documents back in batches, and have been turning them into a structural root-cause analysis. The goal at the end is a publication-ready piece plus a journalist who’ll take the story further. AI is the analyst; I’m the editor and the one whose name goes on it.
A 1920s house in town and a cabin a few hours away. Two sets of vendors, two insurance policies, two utility stacks, two property-tax cycles. The AI doesn’t run the vendors — vendors are humans. The AI runs me. The unlock is “nothing slips,” not “nothing manual.”
Small fusion-dance scene. The work splits into three things: keep regulars regular, win lapsed people back, and make sure the next event isn’t a surprise to anybody. Most of the levers are personal outreach, not marketing. AI helps where the lift is “look at all of this and tell me who to text.”