Pocket Fishing Guide: live data in, AI drafts, a person decides.

Context

Arkansas anglers decide where and when to fish from scattered sources: USGS gauges, National Weather Service forecasts, Army Corps of Engineers lake levels, and the Arkansas Game and Fish Commission weekly report. None of them talk to each other, and none of them say what it means for a given species on a given day.

Corey's role

Founder and sole builder: product, data pipeline, model, front end and operations. There is no team behind this number. That is the point of the case study and also its limit, see below.

What was built

• A data layer that pulls USGS, NWS, USACE and AGFC weekly reports, with Open-Meteo and USNO, refreshed every 20 minutes.
• Water temperature, barometric pressure and spawn phase for 57 waters and 33 species, presented per water on a phone.
• A multi-agent AI architecture that drafts content and code; a person reviews what ships. Every change is a pull request.
• A free, public web app on a static host, with no app-store dependency.

Result

723 users in the first 90 days after launch. The app is free, so there is no revenue figure to report, and there is no baseline because it replaced nothing that existed before.

What carries over to client work

The shape is the same as the workflows set up for clients in the 90-day engagement: public or in-house data in on a schedule, AI drafting, a named person approving, and a runbook so it keeps running when attention moves elsewhere. The lab is where that pattern is proven before it is sold.

What this does not prove

• It is not a client engagement, so it says nothing about adoption inside someone else's team.
• The numbers are first-party, from the app's own analytics, not audited by a third party.
• Usage is seasonal and regional. Treat 723 as a signal that the pattern works, not as a forecast for your business.

More on the lab and what came before it, RaftUp and the NMMA water-taxi app, on the lab page.