been working on dental CBCT segmentation for a while now finally putting it out there. it's a deep learning approach to automatically segment teeth and jaw structures from 3D scans. not glamorous, but honestly kind of cool when it works.
https://t.co/gvaiOGGD9k
AI shouldn't get amnesia every session.
Built a memory arch that works like humans - episodic, contextual, behavioral. Toggle session memory to tune how much context it carries. Full recall or laser focus, you control the weight
๐https://t.co/iJTzlojupo
Any video of a human doing a task should be enough to teach a robot.
I am building the compiler that makes that possible SkillCompiler takes a raw video of a human doing a task and compiles it into robot-ready artifacts: object tracks, segmentation masks, contact events
Anthropic just released Claude Opus 4.8.
Honest take the last 3 models have felt almost identical. Most people can't tell the difference in real use.
Mythos was hyped as a breakthrough. Turns out it was just Opus 4.6 on steroids, now rebranded as Opus 4.8