Blog
What Surprised Us in RoboPapers
To sharpen builders' instincts, here are the findings from RoboPapers that keep contradicting what everyone assumes about physical AI.
10 JUL 2026Why Community Robot Data Is Harder Than It Looks (and How to Approach It)
Everyone who fine-tunes a robot policy on public data hits the same wall: a pile of "same robot" datasets that will not merge. We mapped every layer of it on the OpenArm, and built an agent to drive the cleanup with you.
09 JUN 2026Is the Cloud the Safer Place to Run Your Robot?
The common path for deploying AI models in robotics is deploying them on the edge. When does cloud make sense?
21 MAY 2026While everyone watches for the “ChatGPT Moment in Robotics”, don’t underrate the “Apollo Moment”
An interesting lesson for the robotics ecosystem from how the Apollo space program (almost by accident) seed-funded the semiconductor industry in the 1960s.
11 MAY 2026Announcing Festivus: Physical AI for the Rest of Us
A catalog of robots, datasets, and policies — and a map of which ones actually work together.
06 MAY 2026Pains & Opportunities in the "Hugging Face for Robotics" meme
What people actually mean when they say there should be a 'Hugging Face for robotics' — four open problems we keep hearing about, and an open call to builders.
29 APR 2026The Hardest Part of Shrinking a Robotics Model
Shrinking Pi0.5 by 2.9× on disk and 2.8× faster. The paper gives you the recipe; the pipeline tax is what makes it hard in production.
23 APR 2026My First Week at Haptic: Making a 3B Robot Policy Faster
A one-week POC: fine-tuning Pi0.5 to 100% on LIBERO-Spatial, then distilling its 10-step action head down to a single forward pass with SnapFlow.
14 APR 2026One Click to See the Robot Move
Part 4 of the MuJoCo series. Paste a HuggingFace URL, get a video of the robot moving.
06 APR 2026Where Is the Midjourney for Robot Simulations?
Part 3 of the MuJoCo deployment series. SimGen is a prompt-to-physics-simulation engine.
02 APR 2026The Software-First Robotics Founder
Software-first founders are entering robotics with different assumptions, different problems, and different expectations. Here is what we are hearing.
31 MAR 2026Deploying MuJoCo on Azure ML, Part 2: Microsoft Heard Us
Microsoft's Azure team responded to our field guide. Here's what they said, what we got wrong, and how simup is changing.
21 MAR 2026Deploying MuJoCo on Azure ML: Surprising Pain Points
Every problem we hit deploying MuJoCo on Azure ML, and the fix for each.
06 MAR 2026There Is Plenty of Room (and Pain!) in Physical AI Research
To inspire builders, here is a list of problems that keep coming up again and again in physical AI research papers.
05 MAR 2026Hello World
Introducing Haptic, infrastructure for physical AI labs.