Wanda Wang

@massiviola01 🤩

Pretrained ViTs are powerful but completely static. Run a forward pass a thousand times, and you always get the same result. But what if you could tell the model what to look for? What if you could adapt the representation to encode attributes relevant to your task?

@massiviola01 Great insight 💕

Comparing a new image encoder to VJEPA 2.1 actually says a lot about the future of computer vision. Video features are advancing rapidly, and unified models that ingest both image and video modalities will eventually turn into the industry standard. This area of research is super interesting, and I can't wait to see more papers leveraging this duality in VJEPA 2.1 at the next conferences. For example, it would be cool to learn a task at the image level (e.g., detection or depth) and easily extend it to video with close to zero effort.

@massiviola01 @robbyant_brain Wow amazing 😻

Image foundation models never stop surprising!😮 And this time, it's not the usual big tech players. A few days ago, @robbyant_brain released LingBot-Vision, a new family of vision encoders built natively for dense spatial perception.

AI Engineer Run Club no better way to make friends from Lisbon, London, Italy, Seattle, and many more places @aiDotEngineer @KernelLabs_ai @EntireHQ @lizziepika @FanaHOVA

Just a typical SF weekend: startup rooftop party, YC hackathon judging, a Series C celebration and CEO spotting at Equinox

He didn’t send flowers, he sent a vertically-integrated floral solution with a TAM bigger than my apartment 😍

work hard, play hard @zmzlois

New bling ✨@claudeai