← Back to overview

Web app / live beta

A useful idea, shipped.

Digital products built around real constraints: specify the problem, ship a testable version and improve it against evidence.

Cravate / live beta

PROJECT / 01Live betaCravate app: craving input, cooking equipment and shopping preferencesCravateOne craving. One answer.

Aug 2026 – present · software / product

Cravate

One answer to “what should I eat?” matched to budget, kitchen equipment and nearby shops.

Developed with Claude Cowork using Opus, Sonnet and Fable. A single-file PWA, with a Cloudflare Worker backend and Gemini API. No accounts, no database and zero cost to the user.

PWACloudflare WorkerGemini APIOpenStreetMapFSA data
Open live project

Cravate / development notes

Testing the recommendations

A £6 budget test returned a £7.49 fast-food meal. The response exceeded the budget and failed the cost and nutrition comparison.

The problem

The option to recommend buying a meal had no clear conditions. It also rejected cooking because the original preparation method could not be reproduced exactly.

The changes

Keep recommendations within budget, compare cost and nutrition, and allow simple assembly using prepared ingredients. Match the result rather than insisting on the same cooking method.

Next to test

Whether people find the suggestions useful enough to return. The app is in beta; repeat use still needs testing with real users.

Contact

Get in touch.

Alongside developing JDC Ventures, I’m applying for 2027 entry to mechanical and aerospace engineering degree apprenticeships. I’m open to relevant placements, insight days and conversations about products, prototypes and early-stage ventures.

jdc@jfdcosta.com
LinkedIn ↗