Rhoda AI Breaks Stealth: Jagdeep Singh Teases "Real World" Generalization

For months, Palo Alto-based Rhoda AI has been one of the most anticipated "ghosts" in the robotics industry. Despite having raised over $230 million and reaching a valuation near $1 billion, the company has remained remarkably quiet. That silence ended this week as founder and CEO Jagdeep Singh began a public-facing pivot, signaling that the stealth period for the venture is rapidly drawing to a close.

From Batteries to Bimanual Manipulation

Jagdeep Singh is no stranger to scaling "hard tech" companies. As the former CEO of QuantumScape, he spent years tackling the complexities of solid-state batteries before turning his attention to embodied AI. After maintaining a cryptic "CEO at Stealth Startup" title on LinkedIn for some time, Singh has updated his profile with a new Rhoda AI brand identity.

The update comes alongside a new mission statement of sorts. In a recent LinkedIn post, Singh identified the "robustness gap" as the primary hurdle currently facing the industry.

"One of the most underestimated problems in robotics is generalization from lab to real world," Singh wrote. "A system that works perfectly in the former can fail with minor changes that come with the latter — slightly different orientation, slightly different lighting, slightly different layout."

This focus on generalization suggests that Rhoda AI is aiming to move beyond the highly controlled "lab demos" that often characterize the sector. Singh’s assertion that "machines still aren’t" robust to these changes echoes the industry's ongoing struggle with Moravec’s Paradox—where high-level reasoning is easy for AI, but basic physical interaction remains difficult.

The Box-Opening Benchmark

Rhoda AI also released a brief but technical video demonstration on X, offering the first look at its "general purpose bimanual manipulation platform". The footage shows a pair of robotic arms, equipped with simple grippers, working in tandem to open a cardboard box.

While box-opening may seem mundane, it requires the robot to handle deformable materials and maintain "physical commonsense" regarding force and friction. The use of two arms—a setup common in recent breakthroughs from competitors like Physical Intelligence (Pi)—points toward a strategy focused on complex, human-like dexterity.

Previous reports indicated that Rhoda AI’s hardware is specifically designed for "heavy lifting". Many current humanoid prototypes struggle to stabilize themselves while carrying loads exceeding 50 pounds. Judging by the teasers, it's unclear if Rhoda is positioning itself to solve this industrial-scale challenge.

A Competitive Convergence

The emergence of Rhoda AI adds another heavyweight to a field already crowded by massive funding rounds. The team Singh has assembled—which reportedly includes Stanford professor Gordon Wetzstein and Softbank veteran Vincent Clerc —suggests the company is combining high-end academic research with practical experience in consumer robotics.

By focusing on the "generalization gap," Rhoda is entering a direct philosophical debate with other players in the space. While companies like Sunday Robotics focus on cheap, high-fidelity human data, and Pi focuses on autonomous reinforcement learning, Rhoda’s specific technical path remains to be seen.

When asked on social media when more details would be available, Singh was brief: "More coming soon." For an industry hungry for "factory-ready" tools, that "soon" cannot come fast enough.