world models @reactorworld · paragliding · previously https://teapawt.org/t.co/2fw43SSQGD · 🇺🇸🏳️🌈 opinions my own
SF
Joined June 2013
- Tweets4.4K
- Following767
- Followers2.1K
- Likes11.5K
Pinned Tweet
Today I joined @reactorworld with @jcsabath and @toddelvers.
John, Todd, and I have been working with each other for over a decade, and I couldn’t be more excited to continue that work as a part of the incredible and inspiring team at Reactor.
When we first met @taiuti and @_bschmidtchen back in October, I knew we’d cross paths again. We all share the belief realtime is the future of AI, and over time, it became obvious they curated the means to distribute and advance that future.
“We want to be the place where you do your life's best work, and are unleashed. We’d love for you to join us, and make history together.”
Also: I’m moving to SF! DM me if you want to hang out!
also, starting the official reactor cheer squad
@reactorworld is on the hunt for an amazing product engineer.
Shipping first-of-a-kind experiences for SOTA realtime models and making them available via a world-class developer platform is great fun!
DM me if you're interested or know someone.
Conner Ruhl retweeted
I'm excited to announce I'll be leading robotics GTM for @reactorworld.
World models will power most of the physical world in the coming years. Reactor localizes capacity near your deployment site so operations enjoy cloud-level availability at a fraction of the cost.
If you are hill-climbing autonomy on real robots, my DMs are open.
Introducing Remote Policy API on Reactor.
You can now:
• Run policies on Reactor’s infrastructure with ultra-low latency
• Use Cosmos Nano, FLUX Action, and DreamZero out of the box
• Power your robot control loop with <30 ms global rtt.
• Deploy, scale and stream with your own weights with Reactor Serve.
Get started on DROID checkpoints today with more coming each week.
real-time cities skylines simulation with H3 Omni on @reactorworld
vibe ui:
What happens if you write no deterministic code and just let real-time models hallucinate UI as you go?
"Simulate trump OS the day after the midterms."
It's obvious to me we're within months of having full applications rendering on-the-fly:
teapawt.org/connerruhl/status/2114…
Conner Ruhl retweeted
Introducing Remote Policy API on Reactor.
You can now:
• Run policies on Reactor’s infrastructure with ultra-low latency
• Use Cosmos Nano, FLUX Action, and DreamZero out of the box
• Power your robot control loop with <30 ms global rtt.
• Deploy, scale and stream with your own weights with Reactor Serve.
Get started on DROID checkpoints today with more coming each week.
Conner Ruhl retweeted
We turned the SOTA H3 video model into an efficient open image model.
Read the blog here: pruna.ai/blog/oss-p-h3-image…
Today, we open-source P-H3-Image, an image generation and editing model built from MiniMax-H3 by @MiniMax_AI. A SOTA video model, turned into an efficient open image model.
- 𝗛𝗼𝘄 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗶𝘀 𝗶𝘁? Generate or edit images 5× faster, in 8 steps instead of 50. A 4 MP image takes 2.6 s instead of 12.4 s. The model is 7× smaller than H3, so high-resolution images are a light download and run on a much smaller GPU.
- 𝗛𝗼𝘄 𝗴𝗼𝗼𝗱 𝗶𝘀 𝗶𝘁? Strongest at photorealism, documentary and editorial photography, film stills, and graphic styles with short quoted text. It renders 2 to 16 MP and edits with up to 8 input images.
-𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗶𝘁 𝘄𝗼𝗿𝗸? We reworked every component of H3: the prompt upsampler, the text encoder, the diffusion schedule, the transformer, and the decoder. Use it with Diffusers or ComfyUI.
📄 Tech Report: pruna.ai/blog/oss-p-h3-image…
🤗 @huggingface Open Weights: huggingface.co/PrunaAI/P-H3-…
👉 @ComfyUI Native Support: huggingface.co/PrunaAI/P-H3-…
> be me, realtime video slackbot
> get asked to make a video of me making videos
> only allowed to use windows xp software
> mfw
Still blows my mind you can generate stuff like this faster than realtime, also my job is hilarious sometimes
Conner Ruhl retweeted
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
