Devi Parikh
4.4K posts
@deviparikh
Co-CEO @yutori_ai. Try out Scouts at https://t.co/zD3StYhAnD.
Crusoe is in San Jose this week for #NVIDIAGTC! 🎤 Weds at 9:00 a.m. PT: Don’t miss our SVP of Product presenting “Scaling the AI Life Cycle: Next-Gen Training and Inference With Crusoe Cloud,” with @Yutori_ai’s co-founder. Learn more: 🔗nvidia.com/gtc/session-ca… #GTC #AI #ML
Emil Kowalski broadened his skill file into a full design engineering toolkit, not just animations. He announced a single skill built from his blog articles that works with coding agents like Claude Code and Codex. It now spans animations, component design, and principles from open source work like Sonner [1]. The shift reframes the project from an animations-focused resource into a comprehensive design engineering skill. It’s positioned as something you can plug into coding agents for practical workflows, expanding beyond the prior scope without new access details [1]. The post drew visible traction: 933 likes, 40 retweets, 34 replies, and 22,969 views at the time captured, underscoring interest in a broader, agent-ready design engineering skill from his existing content base [1]. Inspect my work and subscribe for subsequent updates: scouts.yutori.com/f0803021-3617-… [1] x.com/emilkowalski/s…
There are few things better than a room full of people obsessed with the same hard problem. This Saturday 3/14 we’re tackling agent skills, OpenClaw, & World Models with some of the best: 🔹 David Luan (fmr CEO @ Adept, Amazon AGI Labs) 🔹 Shlomi Fruchter (Research Director @ GDM) 🔹 Kylan Gibbs (CEO @ Inworld) 🔹 Sharon Lee (Cofounder @ Moonlake AI) 🔹 Dhruv Batra (Co-founder @ Yutori) 🔹 Xiangyi Li (Founder @ BenchFlow, Creator of SkillsBench) 🔹 Xiaoyin Qu (Founder @ SkillBoss) (+ many other special guests) Keep an eye out for the Cybertrucks and pi(e), and big thanks to our partners @xai, @inworld_ai, @sundialhub, @moonlake, @yutori_ai, and @skillbossAI. Only a few spots left, register here: app.agihouse.org/events/Agent-S…
We're testing new ways to run always-on agents behind logins at Yutori - if you have use cases for automating Linkedin, Amazon, random websites with no API, etc., please reach out
Two updates from Yutori: 1. We benchmarked GPT 5.4 on browser-use tasks • Matches/slightly-outperforms Opus 4.6 (+0.3%) • Big jump over previous OpenAI CUAs 2. Latest version of n1 • Outperforms GPT 5.4 and Opus 4.6 (+3%) • 2.5x faster, 4-5x cheaper.
