RJ Scaringe doesn’t buy the humanoid hype. The Rivian founder says the industry is chasing showy robots that can sprint and flip, when what factories really need are intelligent systems with exceptional hands, simple bodies, and industrial-grade reliability. That thesis anchors his new company, Mind Robotics, which aims to build the AI models, machines, and deployment stack for real production floors.
Mind Robotics has raised a $500 million Series A co-led by Accel and Andreessen Horowitz, valuing the startup at roughly $2 billion. The company has secured $615 million to date. It is a separate, privately held venture that could one day serve Rivian but is designed to stand on its own, with a board that includes Scaringe, Jiten Behl of Eclipse Ventures, an Accel representative, and a Rivian appointee.
The idea surfaced as Rivian mapped out manufacturing for its mid-size R2 program. Scaling to multiple plants meant billions in capex, and Scaringe grew convinced that “classic” industrial robots would endure, but that a new class of robots with human-like dexterity—not human-like gymnastic skills—would be essential.
A Factory-First Rebuttal To Humanoid Hype
Scaringe argues most general-purpose projects copy our bodies too literally. In his view, biomechanics that impress on stage add complexity, power draw, and failure modes with little payoff in a line-side cell. Factory work doesn’t demand vaulting over obstacles; it demands precise manipulation, durable uptime, and safe operation around people.
He also points out the operational design domain matters. Homes are chaotic—stairs, rugs, pets, toys—while factories are structured, mapped, and repeatable. That predictability lowers deployment risk and shortens the path to a usable data flywheel for training models. The target, he says, isn’t “do everything a human does,” but “do the high-value tasks humans do with their hands.”
Hands Over Acrobatics in Industrial Robotics
Mind’s design principle is blunt: everything on the robot exists to put its hands in the right place. That means focusing capital and engineering time on end effectors—grippers, wrists, and tactile perception—rather than on exotic leg mechanics. One module might clamp a four-inch steel pipe with high torque; another might delicately start an M4 fastener into an aluminum housing. Those are radically different demands and likely require different “hands.”
This runs counter to the one-robot-fits-all narrative. In biology, form follows function: the best swimmer doesn’t look like the best sprinter. Likewise, Scaringe expects a family of manipulators optimized for categories of jobs, not a single anthropomorphic hand that pretends to do everything. The approach borrows from the industrial world—where specialized grippers from firms like Schunk or OnRobot dominate—while layering in modern vision-language-action models and high-fidelity tactile feedback.
Models Hardware And The Missing Flywheel
Mind plans to build the full stack: perception and policy models, the mechatronics, and an at-scale deployment platform. Scaringe is skeptical of depending on startups that have never industrialized products, lack supply chains, or have no real data pipelines. The advantage, he says, goes to teams that can close the loop between field performance, simulation, and rapid updates—week after week—on the same workcells.
There is appetite for that. The International Federation of Robotics reports record industrial robot installations in recent years, with China leading and automotive and electronics lines accounting for the bulk of demand. Major manufacturers are already piloting general-purpose systems, from BMW’s work with Figure to Amazon’s trials with Agility Robotics. But much of the sector still leans on traditional arms and bespoke grippers; turning manipulation into a scalable software problem with modular hardware is the unclaimed prize.
Design That Plays Nice With People on Factory Floors
Another contrarian streak: aesthetics and interaction. Scaringe wants robots that look approachable, not “lean and menacing.” The UI must earn trust with line operators who will work beside the machines for years. That implies clear intent signaling, natural language interfaces, and safety features aligned with standards such as ISO 10218 and ISO/TS 15066, while avoiding cartoonish cues that invite pranks or undermine authority on the floor.
Energy efficiency and serviceability are also baseline requirements. In factories, margins die by a thousand cuts: a power-hungry actuator here, a hard-to-swap gearbox there. Simple locomotion, fewer degrees of freedom, and robust hands can lift mean time between failures and shrink total cost of ownership—metrics that matter more than viral demo clips.
Why Rivian Cares About Factory-Ready Robotics
Rivian’s next growth chapter hinges on bringing the R2 to market efficiently. If volumes climb, new plants will follow, and Scaringe doesn’t want to pour billions into lines that go obsolete mid-build. Mind could de-risk that by delivering robots that drop into existing human-centric spaces, integrate quickly with brownfield infrastructure, and get better the more tasks they perform.
For the record, Mind wasn’t always “Mind.” The internal codename was Project Synapse—a nod to the company’s brain-first focus and, Scaringe jokes, a playful tribute inspired by his kids’ school. The name changed, the emphasis didn’t: intelligence at the top, mastery at the hands, and only as much body as the job requires.
The Stakes And What To Watch In Industrial Automation
Mind’s near-term credibility test is whether it can land pilots that beat incumbent automation on cycle time, quality, uptime, and total cost—without heroic engineering at each site. Watch for a library of swappable hands, fast retraining across like tasks, and a deployment cadence that compounds the data advantage. If Scaringe is right, the next breakthrough in robotics won’t look like a sprinter. It will look like a great pair of hands that never gets tired.
