✳flâneur — a map of the web's best reading

francesco215.github.io/autoregressive_diffusion/

francesco215.github.io · 3,497 words · saved by 1 readers

After leaving my job, I was officially unemployed, but don't worry, I had plenty of time to train some models since I'm already training to be a disappointment. Originally, I wanted to build an AI for drones that's as smart as a bee. From first principles, this seemed doable: bees have tiny brains, and with just a brief exposure to the real world they can navigate reliably without ever getting lost. In ML terms, this problem should require little compute and little data (in principle!). But I needed an environment to train the RL policy in. Smashing real drones wasn't an option, and there were no good off-the-shelf long-context world models to train in. So I had to make one myself. In this blog post, I'll talk about building that virtual environment. There won't be any RL yet- the Bee AI has to wait for now. When I started this project (around February), the hottest world modeling papers were DIAMOND and GameNGen. But both suffered from extreme amnesia, so I decided to try something

Oniris A small story about building a long-context World Model with no money Authors Affiliations Francesco Sacco (POV) None Matteo Peluso University of Zurich Date August 22th 2025 DOI 10.5281/zenodo.16927467 Where it all started: The Humble Bee After leaving my job, I was officially unemployed, but don't worry, I had plenty of time to train some models since I'm already training to be a disappointment. Apis mellifera, commonly known as honey bee, with a custom made VR headset. It allows our little user to learn how to fly without bumping around. A bee brain has less compute than modern smarp

Explore this link on the map →

saved by

related reading