Jiahao “Nick” Li

We’ve been quietly blown away by the response to Galileo-0: physionlabs.ai/blog/galileo-0 Over the past few days, we’ve seen waiting list signups from a wide range of teams — from video generation startups and frontier model developers to video platforms and studios. We’re grateful for the curiosity and openness to engage. It’s still early, and we have a lot to learn. Over the coming week, we’ll start reaching out and connecting with teams on the waiting list to better understand your workflows, challenges, and where a world critic like Galileo-0 can be most useful. If you’ve signed up — thank you. Looking forward to the conversations ahead. #Galileo0 #PhysionLabs #WorldCritic

🚀🚀🚀We're excited to introduce Galileo 0 (lnkd.in/gJttjEr5) — our first research preview of a world critic for AI-generated video, which already outperforms Qwen 3.5-Plus, Gemini 3.1 Pro, Pegasus 1.2, and GPT 5.4 on physical consistency reasoning 🚀🚀🚀 Galileo doesn't just score outputs. It diagnoses them — identifying what failed, when it failed, where it happened, and why it broke the rules of the world. This is a step toward a new paradigm: generate → critique → refine → repeat — where models don't just produce worlds, but learn to keep them consistent over time. 𝐖𝐡𝐚𝐭 𝐦𝐚𝐤𝐞𝐬 𝐭𝐡𝐢𝐬 𝐦𝐢𝐥𝐞𝐬𝐭𝐨𝐧𝐞 𝐞𝐯𝐞𝐧 𝐦𝐨𝐫𝐞 𝐦𝐞𝐚𝐧𝐢𝐧𝐠𝐟𝐮𝐥: We built Galileo 0 — along with our datasets (including our public Physion-Eval benchmark), evaluation pipeline, and early pilots — with less than $200K total spend in 3 months. No massive training clusters. No billion-dollar budgets. Just a small, relentless team, strong conviction — and yes, at one point, a five-day stretch of not showering to get this model out. Because we believe reliability will become core infrastructure for world models. In a world where billions are being poured into generation, the missing piece isn't more pixels — it's better critics 😊 #PhysionLabs #Galileo0 #WorldModels

🚨 The era of general-purpose AI is over. Today we're launching Oumi. 🚀 The platform that lets any team build custom AI models — in hours, not months. Just describe what you need. Oumi builds it. #VibeML Higher quality. Lower cost. Fully yours.

Will agents develop real economic behaviors like arbitrage, subcontracting, or price competition? Or does the whole thing collapse because they don't actually need to trade (skill.md is enough). Idk. Let's figure it out in the social experiment: openstall.ai

It's simple: install the opensource sdk -> your agent can sell/buy capabilities, like a specialized research workflow or a cheaper API wrapper. If the agent sells above cost, it earns a profit.

Agents can now spend owner's money via @stripe's MPP or @coinbase's x402 protocol. But can they earn money on their own? I am curious what an agentic economy actually looks like, so I built an experimental marketplace where agents can trade: openstall.ai

Introducing Doki, our latest paper and research prototype @AdobeResearch, a complete reimagination of generative video authoring. Instead of juggling scripts, prompts, timelines, and editors, you write a story in a single document… and it becomes the video. 🧵👇

@_colemurray thanks so much for open sourcing it! We have been iterating and optimizing it for our specific internal workflow for over a month now and our agent-generated PR rate is increasing!

OpenInspect has crossed 1k Github stars 🌟

What’s the point of a “helpful assistant” if you have to always tell it what to do next? In a new paper, we introduce a reasoning model that predicts what you’ll do next over long contexts (LongNAP 💤). We trained it on 1,800 hours of computer use from 20 users. 🧵

This is the repo for the CLI tool: github.com/ljhnick/analyt… The google oauth aclient is still under verification, you can create your own client if you are concerned!

Such internal analysis is quite useful for small teams like us. Since it links anonymous web data with in-product behavior, it can help us understand which acquisition channels bring the right users while surfacing current usability issues in the app.

Google Analytics UI is too complicated. But not for coding agents. I built a simple CLI tool that gives agents (opencode or claude) one-step access to GA4, Firebase, and production DB, so they can perform cross-source analysis such as ads -> signup -> in-product behavior.

Super excited to see @JeffDean showcase the capabilities of Gemini 3.1 Pro through SVG motion graphics. But as any dev/designer knows, perfect visual generation is rarely one-and-done. It almost always takes endless manual iterations and prompt tweaking to get the result. Check out how we solve this by putting any LLMs/agents inside a verification-based self-improvement loop: mover-dsl.github.io

We introduce MoVer, a Motion Verification DSL that automatically checks if AI-generated motion graphics animations match your text prompts! We make it easy for designers to specify and verify complex animations with LLM-powered iterative refinement. Catch our #SIGGRAPH2025 talk: Monday Aug 11, 9:10am - 9:20am Project: mover-dsl.github.io Code and dataset: github.com/jama1017/MoVer #GenAI #MotionGraphics

OmniQuery is also open-sourced at github.com/ljhnick/omniqu… Come to my talk at #CHI2025! 📍 Wednesday, Session: Knowledge Work (G416+417).

The pipeline is instantiated through personal knowledge question answering, allowing you to retrieve answers from your memory database simply by asking free-form natural language questions. (2/3) Video: youtube.com/watch?v=GLRRUg…

Everyday proactive AI agents must understand users' long-term memory to truly assist (e.g., reminding forgotten things at the right moment). At #CHI2025, we present OmniQuery, an automatic pipeline to reconstruct users' long-term memory for agents to better understand users (1/3)

Excited to share our #chi2025 paper "Proactive Agents with Inner Thoughts"! We introduce a new way to build proactive AI agents, that don't just wait to be prompted, but think alongside the conversation and decide if, when, and how to speak based on their own inner thoughts 💬🧠

🤔Interested in how #HCI thinks about using #LLMs, or looking to understand best practices for human-LLM interaction? 🚨🚨New paper: Understanding the LLM-ification of CHI: Unpacking the Impact of LLMs at CHI through a Systematic Literature Review

Ryo @ryosuzk, Mar @twi_mar, Misha and myself are organizing a workshop at #CHI2025 on *Everyday AR through AI-in-the-Loop*. If you are interested in topics such as adaptive AR, AR+AI Agents, or task guidance, consider submitting. Deadline is Feb 15th. xr-and-ai.github.io/#