They can solve the model maybe. But who's solving the deployment, the fleet ops, the hardware failures? Intelligence without infrastructure has no effect on the real world.

funny how "AGI will solve robotics" turned into "we need to hire SLAM engineers"

VR didn't fail. it was just early. the metaverse will probably happen. but we got the sequencing wrong. everyone tried to start with entertainment and social. gaming, virtual offices, hanging out in digital worlds. none of it had enough pull to justify wearing a headset for

Every robotics company we work with has more data than they can use and less infrastructure than they need

the modern robotics infrastructure stack: - simulation and testing (NVIDIA Isaac, Gazebo, MuJoCo) - connectivity and comms (Wi‑Fi/5G/LTE/Starlink VPN backhaul) - fleet management and monitoring (Formant, Open‑RMF integrations) - hardware abstraction and drivers (ROS 2,

the team that co-invented VLAs just abandoned them what this means for robotics teams: 1. if you have enough data, you don't need to inherit someone else's foundation model. clean sheet wins. 2. the VLA approach existed because there wasn't enough robotics data. that

your next piece of furniture is going to be a robot

the gap between 0→1 and 1→n in robotics is where most companies die. the skills that get you your first customer are completely different from the skills that get you your 30th.

owning the best robotics dataset in the world means nothing if you can't learn from it. to learn from it: a) it has to be searchable, fast b) you need to reason about it & find patterns across missions, robots, and conditions c) it has to inform your code, your models, and

if you are a robotics company and want to yell at us weekly lmk

everything a robotics company needs to go from 0→1: - one real customer problem with hard ROI, not a demo - one hero workflow you own end-to-end (automation + human-in-the-loop fallback) - simulation for fast iteration (NVIDIA Isaac, MuJoCo, Drake, Genesis) - a single testbed

this is playing out in robotics faster than anyone expected ... agent-native, not AI-native. we built agents that don't wait for questions. they run scenarios on every new mission automatically. surface anomalies the team missed. flag regressions across sprints. ping Slack

Tesla says 56% of Optimus's cost is actuators. America manufactures almost none of them. Generational opportunity for builders in this space.

Alien Dreadnought, Phase 1: The factory builds the car. The car drives itself out. The car will never be driven by a human.

we are so very early in robotics. the people starting now are going to look like the people who started SaaS companies in 2009.