HackYeah 2026 ReviewDriveMate
MichaelDev · indieedevsh-ui/DriveMatee
Council score
Median of 3 models, weighted by the task's official criteriaA seriously built and well-designed Polish in-car assistant with real code and real urban-data touches, but as submitted it is a personal consumer driving app rather than a Smart City tool, its core voice-AI promise is unverifiable from the evidence, and the required presentation was not delivered with the entry.
The voice-to-real-actions insight and the street-history narration over GUGiK PRG, TERYT, OSM and Wikipedia are genuinely thoughtful, but the core concept converges with CarPlay and Siri and the differentiating AI layer is unproven. Member C's 3 underrates it; the evidence supports the middle view shared by A and B.
Members disagree here: scores range by 2.5 points.Two of three members score 2 and the evidence backs them: navigation, parking and fuel come from Apple Maps search, and the one real urban data call feeds street trivia rather than city operations. Even A's own notes concede a single municipal touchpoint, so A's 5 is the unsupported outlier.
Members disagree here: scores range by 3.0 points.There is a clear user and a genuinely useful one-screen driving scenario with safety-minded details, but nothing verifies it runs: demo checks are empty, the team admits the AI is incomplete, and it requires an iOS 27 physical iPhone.
Eighteen screenshots and a 20-slide deck show a distinctive, consistent liquid-glass cockpit, though the running-app shots reveal placeholder map data, so the built UI trails the concept. C's 5 is the outlier against A and B.
Members disagree here: scores range by 2.0 points.About 13.5k lines of real Swift across 26 services with live MapKit routing, GUGiK and Polish speech integrations, but zero tests and no runtime evidence. A and B's reading matches the measured codebase; C's 3 rests on an unproven suspicion of undisclosed AI-generated code.
Members disagree here: scores range by 3.0 points.Strengths
- Substantial real code, about 13.5k Swift lines across 26 services: MapKit routing with turn-by-turn guidance, a Polish speech pipeline with wake-word training, and camera capture.
- Genuine open urban-data integration in PolishStreetKnowledgeService: GUGiK PRG OGC API, TERYT codes, Nominatim OSM and Wikipedia.
- A distinctive, consistent liquid-glass landscape design carried across 18 screenshots and a 20-slide deck.
- Safety-first UX discipline: overspeed voice warnings, music ducking, and honest refusals instead of invented answers.
Weaknesses
- It is a consumer driver app, not a Smart City tool: no city feed, no city-staff user, and the only municipal data is used for street narration.
- No runtime proof of the core promise: demo checks are empty, there is no video, and there are zero test files and zero test cases.
- The required presentation was not provided with the submission; the deck exists only inside the repo.
- The team admits the AI functions are incomplete, and the AI path depends on specific hardware such as an iOS 27 physical iPhone and Apple Intelligence.
Red flags
- May not be built for this task: a personal consumer in-car assistant for individual drivers, not a day-to-day city tool; the only municipal touchpoint is street-name narration built on GUGiK data, and no city feed, city-staff role or city-level decision is served.
- The submission itself declares 'Not a finalist' and admits the AI functions are not fully complete, so the headline voice-AI claims go beyond what the evidence or the team's own statements support.
claude:glm-5.3-flash55.070% agreedots-studio/dots-3-note-preview:free49.5100% agreeinclusionai/ling-3.0-flash-sante:free34.050% agreeclaude:glm-5.3–judgeThe council, the evidence pack, the prompts and the queue are all on GitHub. If a review helped you, a star helps other teams find it.