Today's projects
Everything the teams built today. Pitches start at 18:30.
- #01formVegetable market
Nuremberg Market Relocation Decision Tool
What was built: a decision-support tool for where Nuremberg's vegetable market should go from 2029. It measures three options (return to the Hauptmarkt, move to the Lorenzkirche, or use the former Kaufhof) on five criteria from open data: nearby residents, transit access, walking time, retail mix, and how evenly visitors spread across shops. Users can set their own weights or use the priorities of eight stakeholder groups, and the tool tests how robust the result is across 1,000 random weightings. It also screens feasibility and forecasts footfall changes. The decision rule is stated up front, and the result is to move to the Lorenzkirche, if room for about 35 stalls can be secured.
AlmostMadeit - #02formVegetable market
UrbanTwin
We built UrbanTwin, a digital twin of Nuremberg's old town that simulates a full market Saturday at the two candidate sites for the vegetable market, St. Lorenzkirche and the former Kaufhof. It uses real data (OSM streets, 3D buildings, census residents, U-Bahn timetables and current roadworks) to show how a senior with a rollator, a vendor with a delivery van, a lunch-break commuter and a nearby shop owner reach each site. It helps the city council decide on facts instead of guesses: every site gets a fair score per stakeholder group, Claude explains the trade-offs in plain language, and simulated residents "speak" about their trips. Most of all, it helps the people a relocation could leave out, such as elderly visitors, market traders and local shops, by showing their problems before the move and testing fixes like a loop stall layout or a U-Bahn kiosk.
Ask Claude - #03formExpress buses
ExpressBus AI
ExpressBus AI is an AI-powered operations platform that helps VAG/VGN identify opportunities to launch additional express bus services despite driver shortages. Using real VGN GTFS timetable data and simulated operational data, the platform detects unused driver capacity, optimizes driver and bus allocations, and recommends new express routes without requiring additional staff. We also built an Emergency Recovery Engine that automatically finds the nearest available bus and driver when a breakdown occurs, helping maintain service continuity. An AI Dispatcher Copilot allows operations managers to ask questions such as "Can we launch a new express route tomorrow?" and receive instant recommendations. The solution is designed for VAG/VGN operations planners and dispatch teams to improve driver utilization, fleet efficiency, and passenger service while reducing operational disruptions.
Team INNEXClaude API, Python, FastAPI, Google OR-Tools, React, Tailwind CSS, OpenStreetMap, GTFS (VGN Open Data) - #04formExpress buses
Nuremberg Public Transport Planner
How can AI help optimise timetables, routes, rotas and available resources so extra express buses can launch with as little extra staff as possible — or with the staff already there? Think optimisation that brings demand, passenger flows, timetables, vehicles and staffing together to surface the most efficient operating scenarios.
BusfahrerClaude Code, OpenStreetMap, VAG API - #05formExpress buses
ZeroStaff-Express
Using optimization for bus routes, free spaces and demand
ZeroStaff Express - #06formExpress buses
Expressfit
We build a strategy and simulator to solve limited bus driver issues in Nürnberg.
GhostdriversClaude - #08formAI for Good
Briefklar
Everyone knows the feeling: an official letter arrives, you read it three times, and still don't know what to do. Now imagine you're new to Nuremberg. Briefklar takes that burden off your shoulders. Take a photo, and it tells you in plain words what the letter means, what to do next and by when. Every answer is grounded in official government sources, not in random pages from the internet. Still unsure? Just ask, and it answers your follow-up questions. Your personal details stay private. Other apps stop there. Briefklar shows the city where readers get stuck, so it can fix a letter once instead of explaining it a thousand times.
BriefklarVue.JS, Python, Claude API - #09formExpress buses
ExpressNet
We built a planning tool that shows where many people in Nuremberg live far from the U-Bahn and S-Bahn, and plans express buses from there to the nearest rail station. The drivers needed come from barely used bus lines that run parallel to the rail network: there, an AI thins out the service frequency without dropping below the minimum standard, and explains every suggestion. Rural lines end at the nearest rail station instead of running into the city.
CodeCampPython/FastAPI backend with PuLP optimisation and a React frontend with MapLibre and deck.gl, built on open VGN, Census and OpenStreetMap data - #10via MCPAI for Good
The AI dividend has no owner
Not an app: a 2-minute talk plus an evidence page. It argues that building AI for Nuremberg is now cheap and what's missing is a named owner. The city's own documents show its Wohngeld AI works (~1.5% error in testing) while waits stayed months-long. Feedback wanted: city people, which of tonight's prototypes would you put your name on, and what stops you?
MaciejClaude Code (Opus 5.5) with Sonnet research subagents and red-team reviews; HTML artifact - #11formExpress buses
Busplaner Nürnberg
We built a web app that creates a duty roster for Nuremberg city buses from the real VGN timetable, legal driving and rest times, and live delays from the VAG, and replans instantly when drivers drop out.
ShiftHappensNext.js, TypeScript, Supabase, VAG-Echtzeitdaten, GTFS - #12formExpress buses
Express from Idle Time
We built a planning tool that finds express buses inside driver time VAG already pays for. It reads the VGN timetable, measures paid idle time at each terminus, and a solver (OR-Tools CP-SAT) fits skip-stop express trips into those gaps while keeping the legal driver breaks. On a normal weekday that's 230 express trips on lines 67 and 68, each about 13.6 minutes faster, with zero extra drivers. Planners check the result in a Streamlit dashboard and can ask Claude where any number comes from. It helps VAG's planners, who've cut bus lines for lack of drivers every year since 2022, and the people who ride the long cross-town routes to Fürth and Langwasser.
Team AIxpress - #13via MCPVegetable market
MarktEntscheid: Wo funktioniert Nürnbergs Gemüsemarkt?
Seit dem 14. September steht der Wochenmarkt wegen der Obstmarkt-Baustelle in der Königstraße statt auf dem Hauptmarkt. Darüber wird gestritten. Wir vergleichen sechs Standorte: Hauptmarkt, Königstraße, Lorenzer Platz, ehemaliger Kaufhof, Jakobsplatz und Kornmarkt. Grundlage sind ein Fußwegenetz aus OpenStreetMap, Einwohner aus dem Zensus, der VGN-Fahrplan und die Passantenzählungen der Stadt. Ein Strommodell zeigt, welche Straßen und Läden Laufkundschaft gewinnen oder verlieren. Die Gewichtung der sechs Kriterien kann jeder selbst einstellen. Empfehlung: Königstraße stützen, Lorenzer Platz offenhalten. Das sind rund 2.200 Marktbesuche pro Tag statt 1.700 am alten Hauptmarkt, mit Ausgleich für die Straßen, die verlieren. Was gemessen und was modelliert ist, steht auf jeder Seite.
Team OlliPython, NetworkX/OSMnx, GeoPandas, Huff-Modell, Zensus 2022, GTFS, OpenStreetMap. Gebaut von Claude Opus mit Subagenten, gegengeprüft von Codex, orchestriert mit Paseo. - #14via MCPExpress buses
Freiraum für Express: Expressbusse ohne neue Fahrer
Nürnberg hat genug Busse für Expresslinien, aber zu wenige Fahrer. Wir haben aus dem echten VAG-Fahrplan (872 Fahrten auf 5 Linien) Umläufe und Dienste rekonstruiert und mit OR-Tools neu verteilt. Ergebnis im Modell: 14 zusätzliche Expressfahrten pro Werktag, mit gleich vielen Fahrern. Die Dienste werden dafür im Schnitt 7 Minuten länger. Mit 2 oder 5 Fahrern mehr sind es 22 oder 37 Fahrten. Die Demo zeigt, woher die Fahrerzeit kommt, was das die VAG kostet (rund 825 € pro Tag) und was es für die Fahrer bedeutet. Ehrlich gesagt: Dienstpläne und Fahrgastzahlen gibt es nicht offen, die sind als Annahmen markiert. Je nach Dienstregel liegt das Ergebnis zwischen 5 und 27 Fahrten. Mit dem echten VAG-Dienstplan wird daraus eine Zahl.
Team OlliPython, OR-Tools (CP-SAT), GTFS (DELFI/gtfs.de), Zensus 2022, OpenStreetMap. Gebaut von Claude Opus mit Sonnet- und Haiku-Subagenten, gegengeprüft von Codex, orchestriert mit Paseo.