🚀Introducing Hierarchical Reasoning Model🧠🤖 Inspired by brain's hierarchical processing, HRM delivers unprecedented reasoning power on complex tasks like ARC-AGI and expert-level Sudoku using just 1k examples, no pretraining or CoT! Unlock next AI breakthrough with

Large language models aren’t replacing traditional browsers anytime soon, but they have become another responsibility for brands on.wsj.com/4lHRTpF

BREAKING 🚨: Comet browser is the first AI assistant which can distribute itself and onboard more users to install it! More Comet invites below ☄️

OK, Perplexity’s Assistant in the new Comet browser is good. Really good.

Perplexity x Tatiana Moodboard #7! The code: --p kuwwd66 The link is in the thread. A sref combo we used to create these images: --sref 257047628 --profile l3h4vio --sw 500 --stylize 500

It’s surprising there’s not more breakout apps per year due to AI coding tools. Outside those tools, have there been any?

Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains 'We introduce Rubrics as Rewards (RaR), a framework that uses structured, checklist-style rubrics as interpretable reward signals for on-policy training with GRPO. Our best RaR method yields up to a relative

Really enjoyed “Apple in China” by McGee.

Clause Code removes the alpha on crafting evals — it’s fun now.

Looking for a senior frontend fullstack hacker to join us on Comet Assistant. DM me.

perplexity.ai/b/zen (on Comet)

I partnered with Comet to try out their new AI powered web browser, Comet I had it plan me a route from Kentuckey to Texas making sure to stop at every Costco along the way for a hot dog

Qwen3-Coder-480B-A35B Qwen vibe coding with Comet in anycoder

Who is I doing video understanding as an API?

Perplexity Comet is insane. I just asked it to find all my subscriptions in Gmail, then told it to nuke all the ones I don't want. Game changer product. This is just scratching the surface.

never take progress for granted. progress takes heroes. every day. if we don't fight for it, we don't just go backwards, we actually lose the abilities of our ancestors.

> Imagine you can train LLMs not from the internet, but from a purely hand-crafted text console game. This is why I’m skeptical about vision reasoning beyond behavior cloning.

We'll continue to see more small Chinese labs releasing impressive models and frameworks, because they build on eachother. This approximates the traditional US-style distributed ecosystem, which was co-opted by hyperscaler aggregation. Eventually distributed problem solving

I wish $0 existed in IDE terminals to refer the the currently selected tab/file.

If you want US to win AI you should ask yourself why we don’t have 100 Deepseeks.