Inspiration

I'm a student in Korea who enters hackathons and contests from all over the world. Deadlines arrive as "5:00pm EDT", "11:59 PM PT" or "UTC", and every contest has its own list of things to submit. This very hackathon closes Oct 26, 5:00pm EDT, which is 6:00am on Oct 27 in Korea. Mixing those up means missing the contest.

What it does

Deadline Radar turns a pasted contest announcement into:

the deadline converted to your local timezone, with the original phrase shown so you can verify it
a live countdown that turns green, amber, then red in the final 24 hours
an editable submission checklist (demo video, public repo, description...) extracted from the text
a list of tracked contests sorted by the nearest deadline, saved in the browser

No backend, no API keys, no uploads.

How I built it

I started from an empty folder with the Devpost Learn skill pack in my coding agent: the skills interviewed me, produced scope.md, prd.md and spec.md (in devpost/), then built the app step by step. Stack: Vite + React + TypeScript, rule-based parsing (regex plus a timezone abbreviation map with daylight-saving rules for ET/PT), Vitest for 39 parser tests, and a browser check of the full flow.

Challenges

Telling a wall-clock time apart from an absolute instant: "ET" follows daylight saving on the contest date, while "EDT" is a fixed offset. A phrase like "11:59 PM PT" with no date can't be guessed, so the app asks the user to confirm instead of hiding the uncertainty.

What I learned

Planning before coding, by letting the AI interview me, kept the scope to one thing that works end to end. Making uncertain parsing visible and correctable matters more than making it look automatic.