Founder and product owner: positioning, formulas, product decisions; agentic development
Status
Live, with a public demo
455
commits in the first three weeks
~2,800
test cases, 139 test files
10
AI coding sessions at once, one repo
11.3M
archived frames in the demo world
For
Team leads first: people who need to read the room, in a place an HR survey cannot reach.
The problem
Ask a team how it is doing and everyone says fine. Surveys are slow, they are performances, and by the time the results arrive the storm has passed.
weathr asks for one tap and shows the aggregate as weather, something anyone reads at a glance, and no one can perform for.
What I did
Owned the product: positioning, the formulas behind the sky, and every decision, recorded as a numbered decision log in the repository.
Had the formulas written as a pure domain package, one source of truth for every formula, shared by the app and the API.
Shipped the web app, a NestJS API and a marketing site, on one box behind one origin. The Android app and the Chrome extension come from the same codebase and are in development.
Ran up to ten Claude Code sessions on the same repository at once, and built the tooling that made that safe. See the agentic-development section of Skills.
The decision that mattered
No language model writes a reading. The sky is computed by formulas anyone can audit; an LLM may one day narrate, off by default, but it may never phrase what the data says.
For a product about how people feel, being able to explain every pixel of the sky is the feature.
The public demo replaying a month of Nevada’s sky, day by day.
Screens
A group’s sky: everyone’s mood, as one forecast.The history, day by day, with the trend for each.
weathr.ai
