Text to 3D simulation on a map (does history pretty well)
Simulate anything on a map from a text prompt -- and conduct risk analysis against LiveUA map's global realtime data points from social media and news sources. I trained a GPT-2-size model on historical incident data used to predict things that will go wrong. As historian Benjamin Breen mentions, the leading language models are good historians, so the application will simulate historical events pretty well also. I include a Multi-Agent RL Urban Mobility model in progress displayed on the map as small white cubes representing traffic and pedestrians. Around SF, it uses real census data and…
In plain words
Text to 3D Simulation on a Map converts text prompts into interactive simulations displayed on global maps. It uses a custom language model trained on historical incident data to predict potential risks and conduct analysis against real-time data from news and social media sources. The tool includes a multi-agent reinforcement learning model that simulates urban mobility, pedestrian movement, and traffic patterns, represented as interactive elements on the map. It currently works with census data around San Francisco and is designed for users who want to visualize scenarios and understand potential outcomes in specific geographic locations.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Simulate anything on a map from a text prompt -- and conduct risk analysis against LiveUA map's global realtime data points from social media and news sources. I trained a GPT-2-size model on historical incident data used to predict things that will go wrong. As historian Benjamin Breen mentions, the leading language models are good historians, so the application will simulate historical events pretty well also. I include a Multi-Agent RL Urban Mobility model in progress displayed on the map as small white cubes representing traffic and pedestrians. Around SF, it uses real census data and other sources for semantically meaningful day plans, etc. It will populate where you move the map, albeit a little slowly. This is based on previous work on my GitHub--I hope to connect it to Unreal's city samples project soon. The simulations are pretty simple so far but will grow in complexity soon. I won the AGI House World Models Hackathon with this and the MARL model. Many thanks to Shota Matsuda and Garrett Johnson for the cloud and atmospheric effects libraries on Github at takram-design-engineering/three-geospatial Glad for feedback, and thanks for trying it out!
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, May 2025
the whole month →
- C9
Life & fun · 2025 · felixrieseberg.github.io
