Lore

Inspiration

Cities are becoming increasingly lonely. Research has linked chronic loneliness to health risks comparable to smoking, and countries like the UK and Japan have even appointed ministers to tackle the problem. Yet while modern cities invest billions into smarter traffic systems, energy grids, and infrastructure, almost nothing is built to preserve the human stories that give those places meaning. We started asking ourselves a simple question: What if a city could remember the people who built it? That question became Lore.

What it does

Lore is a digital memory layer for cities. Residents can pin personal stories to locations on a map, creating a living archive of experiences tied to real places. Anyone exploring the city can discover memories from decades ago, learn the history behind an ordinary street corner, and see places through the eyes of the people who lived there.

Each neighborhood also has a Lore Report: an AI-generated literary portrait written from the memories people contribute. Instead of summarizing events, it captures the emotional identity of a neighborhood and turns collective memories into something worth reading.

How we built it

Lore was built with Next.js for both the frontend and API routes. We created a fully custom SVG city map with deterministic building generation instead of relying on external mapping services, giving us complete control over the experience. AI-generated Lore Reports are powered by the Groq API running Llama 3.3 70B, and the application is styled with Google Fonts to reinforce the calm, editorial aesthetic.

Challenges we ran into

Our biggest challenge was making the AI feel meaningful rather than gimmicky. We didn't want the Lore Reports to sound like summaries or generic chatbot responses. It took a great deal of prompt engineering to consistently generate writing that felt poetic, specific, and grounded in the memories themselves.

Designing the map presented another challenge. We wanted it to feel like a living city without depending on external map APIs, so we built our own procedural SVG-based city from scratch.

Accomplishments that we're proud of

We're especially proud of how the Lore Reports turned out. They genuinely feel like short pieces of literature instead of AI-generated summaries.

More importantly, the core experience works exactly as we imagined. Every memory added becomes part of the city's evolving story, making the map feel more alive with every contribution. The overall experience feels distinct from a typical hackathon project and focuses on emotion as much as technology.

What we learned

One of the biggest lessons we learned is that the best AI experiences are often the least obvious. Instead of giving the model complete freedom, we found that carefully constraining the prompt produced much stronger writing.

By limiting each Lore Report to under 110 words, two paragraphs, third-person perspective, and present tense, the output became far more consistent, expressive, and believable. More thoughtful constraints led to better creative results.

What's next for Lore

Our next step is bringing Lore into the real world. We plan to support real GPS-based memory locations so stories can be attached to actual places instead of a fictional city.

We're also building Memory Walks, where AI generates emotionally meaningful walking routes through neighborhoods, along with a live city-wide heat map that visualizes where memories are being shared, revealing the places that matter most to a community.

Built With

• googlefonts
• groq
• llama-3.3-70b
• next.js
• vercel