Building my first app

A course project has to work once, in front of people who want you to pass. An app has to work every time, for strangers who owe you nothing. March was the month I learned the distance between those two things.

The app was SniffCheck — a job-ad scam checker. Paste a job ad or recruiter message and its AI runs a deep scan for predatory language, misleading titles and hidden red flags, then returns a scam risk score out of 10 with every flag explained in plain English. Behind it sits an educational reference of the scam patterns scammers reuse.

Some decisions were about what the app wouldn't do. No account. No data collection. No stored scans. Partly principle — the app is for people at their most vulnerable — and partly a beginner's honest advantage: you can't leak what you never collect.

By the end of March it worked. Paste an ad, get a score, see the red flags explained. It was built as a webapp on Google AI Studio, wrapped with Claude, and pointed at one place: Google Play.

March also brought a hard decision: I gave up on RailRoam. Advice from people I trusted was blunt — a bigger, well-established company could swallow the idea whole the moment it showed promise. So I shelved it. Shelved, but not forgotten.