🎉Our work 𝗧𝗲𝗮𝗰𝗵𝗶𝗻𝗴 𝗮𝗻 𝗔𝗴𝗲𝗻𝘁 𝘁𝗼 𝗦𝗸𝗲𝘁𝗰𝗵 𝗢𝗻𝗲 𝗣𝗮𝗿𝘁 𝗮𝘁 𝗮 𝗧𝗶𝗺𝗲 was accepted to ECCV 2026 !
We taught an agent to draw vector sketches — part by part. Joint work with @RichardXu110572 ,David Yunis, @YVinker & Greg Shakhnarovich
More in thread🧵
I spent a weekend building an interactive atlas of traditional Chinese timber architecture with @OpenAI's GPT-6 Astra in Codex. Its 3D modeling abilities absolutely blew my mind. 🤯🤯
Introducing 木构图志 · 𝚃𝚒𝚖𝚋𝚎𝚛 𝙰𝚝𝚕𝚊𝚜.
Try it here → timber-atlas.xiaodan.io
This
Karpathy’s point about models struggling to audit and interact with what they create really resonates with me. What I’d love to work on is visual agents that can maintain a persistent understanding of an evolving world, reason about the consequences of their own actions, and
We're starting to leave the territory where you'd test an LLM by e.g. "create an svg of pelican on a bicycle". As one idea to generalize it, I was interested what Opus 5 would do if I gave it the first paragraph of the Lord of the Rings, a 1M token budget (~$10) and asked for
🤖✏️We taught an agent to draw vector sketches one part at a time: look, read a instruction, draw, look again, repeat.
New preprint: Teaching an Agent to Sketch One Part at a Time
Joint work with @RichardXu110572, David Yunis, @YVinker and Greg Shakhnarovich.
More in thread 🧵
BREAKING NEWS
The Royal Swedish Academy of Sciences has decided to award the 2024 #NobelPrize in Physics to John J. Hopfield and Geoffrey E. Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks.”