Seaside
A city-scale AI simulation where hundreds of agents reason, adapt, and react to the real world.
What we built
Seaside places 400+ AI agents in the city of Seattle. Each agent has a name, personality, job, home address, and daily routine. Every day, an LLM plans their entire schedule — where they go, what they do, and why — based on who they are and what's happening in the world around them.
Agents commute on real roads using Mapbox routing. They eat at real restaurants, work in real offices, and make decisions the way people do — by reasoning about context, not following scripts.
How it works
1. Scrape the real world
Apify pulls live Seattle news — road closures, weather, events, transit disruptions — from Google News.
2. Store & summarize
Raw data is stored in Box. Box AI reads through the headlines and generates a concise world-context summary.
3. Agents reason & adapt
AWS Bedrock (Claude) plans each agent's day using their personality + the world context. Close a highway? Commuters reroute. Rain forecast? Outdoor plans shift indoors.
4. Visualize & analyze
Watch the simulation play out on a live Mapbox map. Scrub through days, track individuals, compare behavior patterns.
Who this is for
🏙️ Urban Planning
Test infrastructure changes before breaking ground. What happens if you close this street? Add a bus route? Build a stadium?
🏪 Retail & Real Estate
Predict foot traffic for site selection. Simulate where realistic consumers cluster before signing a lease.
🎪 Event Management
Simulate crowd flow before doors open. Find bottlenecks, plan security, optimize vendor placement.
📊 Market Research
Test campaigns on synthetic consumers with real demographics before spending a dollar on media.
Where we're going
Seaside is currently a proof of concept — 400 agents, one city, seven days. But there's no architectural limit. The system is city-agnostic. Swap the bounding box and seed data, and you have Portland, Austin, or Tokyo.
- → Scale to thousands of agents with distilled behavior models
- → Multi-city support with custom demographics
- → Inter-agent interaction (social networks, word-of-mouth)
- → Backtesting against real historical data for validation
- → Self-serve platform for scenario planning
Team
Built by Shaan Shekhar, Priyanshu Mahey, and Rupak Ghosh at the AWS × Box × Apify hackathon in Seattle. Honorable mention winner.
Interested in what Seaside can do for your use case?
We're exploring partnerships and pilot customers.
Get in touch