Make My Circuit

Inspiration

I run a London art community of about 1,000 people. The Circuit already helps us find current exhibitions and follow curated gallery routes. But a fixed route cannot answer a personal question: if I am in Soho at 2pm, have two hours and want painting, what should I see without crossing London all afternoon?

Make My Circuit answers that question.

What it does

A visitor chooses a date, starting area, start time, available time, artistic interest and longest walk. The planner filters the live Circuit catalogue and builds a timed walking itinerary from exhibitions running that day. It explains why each stop fits, maps the route and opens the same walk in Google Maps.

Removing a stop runs the planner again. It does not merely hide the rejected card. The order, timing and walking total are recalculated around a replacement exhibition.

How I built it

The Circuit is a Next.js app deployed on Vercel, with its exhibition catalogue in Airtable. A compact server route sends only the fields the planner needs. The browser-side planning module handles eligibility, preference scoring, route order, walking estimates, visit time and alternatives. MapLibre and OpenStreetMap render the route.

I used Codex with GPT-5.6 to extend the existing product. Codex inspected the current Routes experience and data model before making changes. It helped implement the planner and responsive interface, write deterministic tests, diagnose the live catalogue, deploy the production build and verify the public flow in a real browser.

I made the product decisions throughout. I kept the feature inside Routes, limited this version to art, chose a deterministic runtime instead of adding an API call for its own sake, and required rejection to rebuild the itinerary rather than simply remove a card.

What was already there

The Circuit product, event catalogue, map and curated routes existed before OpenAI Build Week. Make My Circuit, its compact planner data surface, the replanning flow, tests and responsive UI are the new work. The repository includes the baseline tag hackathon-baseline-2026-07-18 for comparison.

Challenges

The hard part was turning live exhibition data into a plan that respects date, time, geography and the visitor's walking limit. The route also had to fit naturally inside a mature mobile product. During final testing I corrected the walking total so it matches the legs shown on the map and made the longest-walk preference a strict limit.

What I am proud of

The feature works with the live London catalogue and is already deployed in the public product. The suggested start time, gallery sequence, map and walking total remain consistent. Replanning produces a real alternative. The feature also has focused automated tests and was verified on both mobile and desktop before submission.

What is next

I want to test the planner with Circuit members and learn which controls actually change their decisions. That evidence should come before adding evening venues or social coordination.

Built With

• airtable
• codex
• google-maps
• gpt-5.6
• javascript
• maplibre
• next.js
• openstreetmap
• react
• vercel