Will be in SF from 9-11 August to attend the AI Coding Agents Hackathon at the Y Combinator offices! Would love to meet folks on Sunday (10th) and Monday (11th) if they're around to chat about 3D Computer Vision, AIGC, and Model Optimization/Quantization. DMs open!

For what it’s worth, I have absolutely no idea how one is supposed to use mmcv. Absolutely none.

Beautiful. Needs to be more common to do this.

Here are tsunami safety tips for those in coastal areas. Boaters may need to leave harbor and go to a depth of at least 50 to 1000 fathoms. People on land should avoid harbors & marinas, get off and away from the beach. #CAWX

It's not public yet but thanks for the support folks!! Do join DSPy discord, I ramble/provide updates there every evening about this :P

An experience of a lifetime. The definition of Cinema is changing.

My hack for learning new concepts from papers (if trying to get a quick high-level grasp), is to look at any downstream papers and see their related works section for a quick and distilled version and/or their method to see if they talk about it. Works like a charm every time.

GitHub Models now offers a free, OpenAI-compatible inference API for every GitHub account. 🎉 No keys. No consoles. No conflict. Let anyone try your AI tools instantly. 🚀 Here's how we're making it easier to build with open source AI.👇 github.blog/ai-and-ml/llms…

Can't even download a dataset without getting rate-limited these days. Sketchfab rate-limiting is crazy. And GitHub will take me too long to download stuff. Can't even download Objaverse without waiting for a few days or something, it seems at first glance

They can’t stop winning, huh

I’m on ChatGPT Pro. How do I access GPT 5? It’s not even showing up on my selector! OpenAI OpenAI Developers

Now that's not very comforting, is it?

Agent Mode is just broken. Keeps saying “Model Not Found”, on app and ChatGPT both.

Why is everyone building tools and not products?

The Bay Area.

I have ~8hrs free in San Francisco & the Bay Area. Who can I meet?

Jules is built for async workflows. Here's how to lean into its strength. Step 1: use AI Studio to write a 'Taskmaster' prompt to turn your goal into a detailed spec: constraints, acceptance criteria, Mermaid flow. Step 2: give it to Jules with your prompt. Come back to a clean

Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful, high-resolution image features. For the first time, a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense

Changing times :) Love this.