x.com/i/article/1950…

okay, let me get this straight… 1/ you can use sora 2 to create hollywood level videos in seconds to build audiences of millions 2/ you can deploy donotpay style ai agents that call comcast, cancel subscriptions, and renegotiate bills automatically 3/ you can clone your

If you've been trying to figure out DSPy - the automatic prompt optimization system - this talk by Drew Breunig is the clearest explanation I've seen yet, with a very useful real-world case study youtube.com/watch?v=I9Ztkg… My notes here: simonwillison.net/2025/Oct/4/dre…

Back to my fine-tuning with MLX on M3 Ultra! Here LoRA with rank 512 on Qwen3-0.6B. Preparing a batch inference provider to make evaluation phase fly 🚀

Hooray! Ladybird has passed the 90% threshold on web-platform-tests! 📈🎉 This is the arbitrary limit Apple says we must reach to be considered an eligible alternative browser engine on iOS. (+ other requirements) So proud of the team for getting us this far 💜

Build a multi-agent home renovation team with Google ADK and Gemini 2.5 Flash Nano Banana. Just upload your room photo, design inspiration and budget to get a complete renovation plan with material, cost, timeline, and an after image of your new room. 100% Opensource code.

Step-by-step tutorial with Opensoruce Code: theunwindai.com/p/build-an-ai-…

I love this line of research from my colleagues at Apple: Augmenting a language model with a hierarchical memory makes perfect sense for several reasons: - Intuitively the memory parameters should be accessed much less frequently than the weights responsible for reasoning. You

what if you could turn two hours of AI work into a week of viral content that gets 2M+ impressions/week? every morning, dan koe opens chatgpt/claude and uses them like an internal strategy team and it's paid off through millions of new followers and $$. he feeds them

OpenAI's prompt optimizer in AgentKit is GEPA. It's gonna be everywhere...

Drew Breunig GEPA mega thread with paper breakdown, videos, and training resources. x.com/jacksonatkinsx…

New Google Research paper builds a personal health assistant that reads a user’s data, answers health questions, and coaches daily habits. It evaluates the system on 10 tasks with 7,000+ human annotations and 1,100 hours from experts and users. The assistant covers 4 needs,

My brain broke when I read this paper. A tiny 7 Million parameter model just beat DeepSeek-R1, Gemini 2.5 pro, and o3-mini at reasoning on both ARG-AGI 1 and ARC-AGI 2. It's called Tiny Recursive Model (TRM) from Samsung. How can a model 10,000x smaller be smarter? Here's how

This is how Anthropic decides what to build next—and it's brilliant. Instead of endless spec documents and roadmap debates, the Claude Code team has cracked the code on feature prioritization: prototype first, decide later. Here's their process (shared by Catherine Wu, Product

The King is back 👀

Y’all really like the Qwen3-VL family of models 🙌🏽 Don’t worry it’s my top priority today 👌🏽

THIS chart is the CLEAREST signal of where the internet is heading. social media time is SHRINKING for the first time in HISTORY, and young people are leading the pullback. Brainrot is OUT. they grew up online, saw the full cycle of social platforms, and learned early that

Amazing roadmap for senior engineers to embrace the tooling.

BOOOOOOOM! CHINA DEEPSEEK DOES IT AGAIN! An entire encyclopedia compressed into a single, high-resolution image! — A mind-blowing breakthrough. DeepSeek-OCR, unleashed an electrifying 3-billion-parameter vision-language model that obliterates the boundaries between text and

I recently spoke to a marketer who ran a $40M brand with just two designers and ONE AI process: He chains 7 different tools together for: ideation, image creation, video editing, and iterating based on performance. ALL using AI and two offshore designers. I’m paying him