uRun

heading to @MLSysConf 2026 next week in Bellevue if you're working on inference, real-time systems, or ML infra, come say hi. always up to trade notes on what "production" actually looks like for stateful, interactive AI workloads.

Every modality in AI follows the same arc: single-shot expensive generations to multi-turn cheap interactive loops. Text and image already went through it. Video is next.

Thank you to the early ones. 🙏 More to come. #WhatCanuRun

Two years ago, the open problem was getting an AI video model to produce a coherent 5-second clip. Recent techniques like Long Live and self-forcing solved that piece. The new bottleneck is serving it interactively. Labs are chasing the next model. The infra layer underneath is wide open.

Real-time interactive video is the hardest workload there is. Every frame has to land inside the 300ms human-perception bar. That's why we're starting there with @urunml. The rest is downhill.

Who we most want building on uRun: creative tooling companies and the studios behind tomorrow's video games. They'll go places we can't imagine → urun.sh #AIvideo #GameDev #VFX

@OpenAI and @Anthropic both charge ~2.5x for "fast mode." The most underrated pricing signal in AI right now.

some snapshots of our launch party @ Joey the Cat in SF last week. skee-ball, open bar, and real-time AI video on every screen. thank you to everyone who came out and pushed the demos somewhere great and weird. #WhatCanuRun → urun.sh

Introducing the founding team with three unique angles on the same problem. Keegan ran inference at Luma during the Dream Machine launch. Sean wrote the O'Reilly book on Docker and has our GPU orchestration dialed in. Matt was running low-latency edge inference at AWS in 2017 (back when "real-time AI" meant the cameras at Amazon Go). We built uRun for the infrastructure bottleneck no one else is solving. urun.sh #AIvideo #FounderStory #realtimeAI #VideoInfra #GenerativeAI

urun.sh launch party - Wednesday, April 29 · 6PM: 🕹️ Arcade games 🍹 Open bar 💻 Live demos 🥽 Meta Quest Giveaway Spots are limited - click the link to grab your invite. 👉 luma.com/3vemq53b

The model moat is shrinking fast. Kimi K2.6 just beat GPT-5.4 and Claude Opus 4.6 on SWE-Bench Pro. But the story isn't the benchmarks - it's the execution layer: → 300 parallel agents → 13 hours autonomous coding → 4,000+ tool calls in one run It's no longer intelligence per token. It's tokens per second. Source: kimi.com/blog/kimi-k2-6 #claude #moonshot #OpenSource

Reminder 🚨 We're going live on Twitch TODAY at 2pm PT. Come hang and bring your questions 👇 twitch.tv/urunml #Inference #Infrastructure #twitch

@faizan10114 @keeganmccallum3 Keep an eye in the coming weeks. Very exciting things on the horizon!

@keeganmccallum3 @urunml cc @keeganmccallum3 inference for interactive, is definitely not what I had considered, a really neat angle. bullish on this. x.com/faizan10114/st…

Today, we come out of stealth. 👋 @urunml is the inference cloud for the interactive era. We wrote down why we're building it and what we believe. Manifesto→ blog.urun.sh Join the waitlist → urun.sh

Imagine exploring this in real time as it generates. That is the infrastructure problem we have been solving. Check us out -> Urun.sh

This is the kind of model we built @urunml for. Real-time interactive worlds need a different infrastructure layer. We have it at uRun.

We just launched @urunml. Come celebrate with us 🎉 Going live on Twitch tomorrow at 2pm PT 🎙️ Covering: → Who we are and why we built uRun → What the interactive era of AI actually unlocks for builders → We might show off something too 👀 twitch.tv/urunml

@NVIDIAAIDev Every era of computing has its infrastructure moment. Generative 3D worlds are going to need one too. The session layer for real-time exploration does not exist yet. We have been building it. urun.sh

Today, we released Lyra 2.0, a framework for generating persistent, explorable 3D worlds at scale, from NVIDIA Research. Generating large-scale, complex environments is difficult for AI models. Current models often “forget” what spaces look like and lose track of movement over time, causing objects to shift, blur, or appear inconsistent. This prevents them from creating the reliable 3D environments required for downstream simulations. Lyra 2.0 solves these issues by: ✅ Maintaining per-frame 3D geometry to retrieve past frames and establish spatial correspondences ✅ Using self-augmented training to correct its own temporal drifting. Lyra 2.0 turns an image into a 3D world you can walk through, look back, and drop a robot into for real-time rendering, simulation, and immersive applications. ➡️ Learn more: research.nvidia.com/labs/sil/proje… 📄 Read the paper: arxiv.org/abs/2604.13036

@zhzHNN We also launched today inference infrastructure built to serve these models in real time -> urun.sh

Today feels like World Model Day. Beyond LLMs, many other fields are still moving fast. With enough capital and brilliant people going deep, real progress keeps happening. x.com/i/status/20444… - NVIDIA: t.co/bC32ADfvDS - Tencent: 3d-models.hunyuan.tencent.com/world/ - World Lab: sparkjs.dev/examples/strea…

@bilawalsidhu The infrastructure question for real-time exploration of environments like this is something we have been deep in. Interesting timing.

Holy crap, NVIDIA just made it drastically easier to create large scale explorable 3d worlds. No manual stitching of smaller 3d generations like other 3d models. Lyra 2.0 looks pretty damn impressive.