SpaceX has exercised the option to acquire @cursor_ai in an all-stock transaction with the goal of building the world’s most useful AI models. For the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon. We look forward to working closely with the Cursor team to advance our frontier AI capabilities

The Eno Genesis launch is pretty 🔥 No head. No face. Zero exposed cables. Just locked in on pure function: Fast Facts • Unmatched Dexterity: 22-DOF hands precise enough to handle sticky tape and lab equipment. • Radical Transparency: A chest screen displays the robot's real-time reasoning before it acts. • Brainpower: Powered by the GENE foundation model. Two thoughts on this 1. I think is the start of reimagining form factor 2. This is just the beginning of “robot labs” rolling out humanoids. Expect many more to come even if they’ve positioned themselves differently publicly prior.

What you work on has never been more important. Make sure it’s fun. Make sure it’s insanely ambitious. That’s the greatest moat.

The embodied AI industry has one huge problem: Data. Robots need massive amounts of real-world experience to learn how to move, interact, and complete tasks safely. To solve this, the industry is building a 3-layer “Data Pyramid”: 1️⃣ Real Robot Data Humans control real robots while every movement, camera feed, and sensor signal is recorded. Pros: This is the best and most accurate data for training robots in the real world. Cons: Very expensive and slow to collect because it needs physical robots, workers, and large facilities. This is why companies like Agibot (Zhiyuan) are building large robot data collection centers in China. 2️⃣ Simulation Data Robots train inside virtual worlds and physics engines. Pros: Cheap to scale. Robots can practice millions of situations safely and quickly. Cons: What works in simulation often fails in real life because reality is much more messy and unpredictable. This is the path NVIDIA is investing heavily in. 3️⃣ Internet & Human Video The internet already has billions of hours of humans doing everyday tasks. Pros: Huge amount of cheap data. Helps AI understand space, movement, and human behavior. Cons: Videos show what humans do, but not the exact force, pressure, or motor control needed to copy the action. Companies like Figure AI, Physical Intelligence, Apple, and NVIDIA are now pushing hard into this area. No company will win using only one layer. The future belongs to those who can combine all three especially whoever turns everyday human video into useful robotic training data at scale.

NVIDIA's Isaac GR00T humanoid (built on Unitree H2 Jetson Thor) needs one thing to learn: real human behavior. PrismaX solves it with VR teleop — put on a headset, control the robot like you're inside it, and every move becomes training data. control → data → training, all in one loop. wild pipeline.

This makes me doubt the 90% success rate that every VLA-based policies claim I truly respect all the work behind public demos, and especially from hardware companies But from what I saw at #ICRA , even pick-and-place is still challenging for VLAs and requires human-in-the-loop

Something tells me we're going to see a lot more of these glossy black oval devices this year.

This is how I’ve run my life btw. I stayed in pc gaming for 14 years because that’s what I cared about and skipped mobile and cloud hype waves. I’ve joined @fdotinc because I love helping others founders and i get a lot of joy from that.