Inspiration
Agentic AI Build Week is busy by design: builders move across venues, workshops, mentors, perks, teammates, deadlines, and Demo Day pressure. During the event, the hardest question is often simple: "What should I do next, right now?"
AABW Next Move was built for that moment. It helps builders act faster, cut confusion, and get more value from the week while they are actually in the room.
What it does
AABW Next Move turns a builder/team profile, current day, current time, event schedule, venues, resources, mentors, teammates, and deadlines into a concrete action plan.
Core workflows include:
- Ranked Next Move recommendation for the current AABW context.
- Event Mode for live venue movement, support routing, and deadline risk.
- Venue map with current and recommended locations.
- Live Triage checklist for venue, mentor, perk, deadline, team, and community blockers.
- Team Radar to find collaborators who cover skill gaps and receive help back.
- Resource, perk, mentor, and deadline matching.
- Copyable action plan, support request, team intro, progress update, and Devpost summary.
- Downloadable calendar hold for the recommended session.
- Saved plans and checklist progress for the current event session.
The product is intentionally not a thin chatbot. It uses AI as part of a structured event workflow with deterministic guardrails for time, venue, and deadline constraints.
How we built it
The app is built with Next.js, React, TypeScript, Tailwind CSS, shadcn/ui-style components, Supabase, pgvector, OpenAI, Leaflet, and OpenStreetMap.
The recommendation engine combines:
- Deterministic scoring for time, venue, profile fit, capacity, and urgency.
- Retrieval over event data using Supabase pgvector when embeddings are available.
- Keyword fallback when pgvector is not configured.
- OpenAI-generated explanations when
OPENAI_API_KEYis configured. - Deterministic fallback so the demo still works without external services.
The project includes mock/public event data so it can run self-contained for judging, while still supporting Supabase-backed persistence and semantic retrieval.
Challenges we ran into
The main challenge was keeping the app useful during a real event instead of becoming another generic AI chat interface. We focused on workflow outputs: concrete next actions, venue movement, deadline guardrails, support requests, teammate intros, and saved plans.
Another challenge was making the app reliable even when OpenAI, Supabase, or embeddings are not configured. The fallback system keeps the product usable offline or in a fresh judging environment.
Accomplishments that we're proud of
- Built a working live-event copilot instead of a static concept.
- Added deterministic guardrails for time, venue, and deadlines.
- Added Event Mode for in-the-room decisions.
- Added Team Radar for builder/team matching.
- Added copyable operational outputs for Discord, WhatsApp, team notes, and Devpost.
- Kept the app self-contained with mock data while supporting real Supabase and OpenAI integrations.
What we learned
Builder experience tools need to reduce decisions, not add more UI. The strongest flow is not "ask a chatbot"; it is "show me the next best action, why it matters, what risk to avoid, and what I can copy/share right now."
What's next
- Add Supabase Auth and per-team ownership.
- Integrate official AABW schedule and venue feeds.
- Add live RSVP/seat capacity.
- Add push reminders for urgent venue moves and submission deadlines.
- Add Discord or WhatsApp bot delivery for support requests and team intros.
- Add turn-by-turn routing between venues.
Built With
- biome
- leaflet.js
- next.js
- openai-api
- openstreetmap
- pgvector
- playwright
- react
- shadcn/ui
- supabase-postgres
- supabase-ssr
- tailwind-css
- typescript
- vercel

Log in or sign up for Devpost to join the conversation.