Stefano Bonato
- Here's some of my ideas that came up in conversation, prototyped fast, for fun.
- Some even get finished. Have a look below.
Done
- a Letterboxd-based movie recommender for me and my gf: our taste, tonight's mood, and what the crowd thinks.
- a way to keep the same coding-agent session alive while hopping between accounts when one runs dry.
Private Alpha
- audio plugins, made by just describing them (furo.re).
- a bilingual, source-checked data site tracking how Italy legislates against femicide.
Being packaged
- Tinder for matching private ideas, without revealing them to the wrong people. Cryptographically, somehow.
- my whole Orange Pi setup: an open-source stack squeezed dry for torrent streaming and home-server chores, all behind Mullvad.
- an Immich extension that sorts recordings by who's in them and cooks up deepfake videos, locally.
In development
- a way to promote coding-agent sessions into tmux, then watch, shuffle and poke at them from a shared, lightweight 2D web canvas.
status: works on my machine. needs real testing. - a first-person roguelite survival game loosely inspired by an old, fucked-up 4chan /vg/ post.
status: lore, physics and audio mostly done. props now, then NPCs and interactions. - a video-editing pipeline that turns raw footage and mic recordings into a short film, brainstorming with you along the way, all on local AI. It retouches, fills in the scenes you forgot to shoot, and re-narrates everything with zero seriousness.
status: chewing through half a terabyte of camping vlogs for the first cut. - a 3D Murano across four centuries, pieced together from piles of loose text and images.
status: waiting for new data from my art researcher partner. - a dark fantasy world made up as you walk through it. A world model underneath keeps it consistent in space and time, and keeps the generation from going off the rails.
status: moving to the cloud. my local rig can't keep up.
Abandoned
- a bedtime story generator that lets the kid butt in and steer the story in real time.
lesson learned: an LLM narrator will go wherever the kid takes it. that's the problem.
My bread and butter: computer vision and machine learning for robots, in Zurich.
Contact: LinkedIn.
