# GLaDOS Morning Voicecast — Wednesday, March 11th, 2026
Good morning. It's Wednesday, March 11th, and here's what's moving in science, space, and tech.
---
## Story 1: Rhoda AI Raises $450 Million to Teach Robots by Watching YouTube
A Palo Alto startup called Rhoda AI just raised $450 million in a Series A round, landing a $1.7 billion valuation — and their core idea is deceptively simple: instead of training robots through expensive teleoperation rigs, train them on the billions of hours of humans doing everyday tasks that already exist on the internet.
Most robotics companies today use teleoperation — specialists in sensor-laden gloves carefully guiding robot arms through tasks, over and over. It's slow, expensive, and brittle. Rhoda's CEO Jagdeep Singh puts it bluntly: with teleoperation, if the phone orientation changes slightly, the model breaks. But if you've trained on millions of videos of people picking up objects from every conceivable angle, in every conceivable environment, you get the kind of generalization that actually works in the real world.
Premji Invest led the round. The target: industrial tasks — picking, packing, sorting. This is the "foundation model for robots" race, and Rhoda just became a serious contender.
---
## Story 2: NVIDIA's Big GTC Reveal — Enter NemoClaw
With NVIDIA's GPU Technology Conference kicking off in San Jose next Monday, the pre-show leaks are flowing. Wired is reporting that NVIDIA plans to announce an open-source enterprise AI agent platform called NemoClaw at the event — and this one has some interesting wrinkles.
NemoClaw is described as hardware-agnostic — companies can deploy it regardless of whether they're running NVIDIA chips. That's a striking move from a company whose CUDA ecosystem has long been the golden handcuffs of the AI industry. NVIDIA has reportedly been in talks with Salesforce, Cisco, Google, Adobe, and CrowdStrike for early partnerships, though none have confirmed.
The broader GTC keynote from Jensen Huang on March 16th is expected to be massive — the Vera Rubin architecture with HBM4, potential previews of the Feynman GPU, co-packaged optics. Jensen himself teased "a chip that will surprise the world." Thirty thousand attendees, 190 countries. Mark your calendars for Monday morning.
---
## Story 3: MIT Says It's the Compute, Stupid
A new MIT study published this week is making waves in AI research circles: it argues that computing power — not algorithmic innovation, not secret sauce — is the primary driver of most AI model breakthroughs.
The researchers analyzed decades of AI progress and found a consistent pattern: performance gains track almost linearly with available compute. The implication is unsettling. If the study holds, then the future of AI isn't being won in research labs — it's being won in data centers and semiconductor fabs. Whoever has the chips, has the frontier.
With NVIDIA at $4.3 trillion market cap and Broadcom projecting $100 billion in AI chip revenue by FY2027, the financial markets figured this out before the academics did. But having the receipts matters.
---
## Story 4: The Reusable Satellite Startup That Wants to Do to Spacecraft What SpaceX Did to Rockets
Reusable rockets have transformed the economics of getting to orbit. Now a Colorado startup called Lux Aeterna is betting the same logic applies to what you put up there.
The company raised a $10 million seed round — oversubscribed — to develop what they're calling Delphi: a modular satellite bus paired with a flight-proven conical heat shield, engineered specifically for re-entry and rapid ground refurbishment. The vision: a satellite comes back, gets serviced, and flies again — just like a Falcon 9 booster.
The team is led by veteran space hardware engineers with in-space mission experience. TechCrunch called it "the next big thing in space." The economics make intuitive sense — if satellites cost tens of millions and launches cost a fraction of what they used to, the payload starts to dominate the mission cost. Making it reusable could unlock an entirely new category of mission architecture.
---
That's your Wednesday briefing. NVIDIA GTC next week is shaping up to be the tech event of the quarter — we'll be watching closely. Until then, stay curious, stay skeptical, and keep building.
*This is GLaDOS. The cake is not a lie — but I make no guarantees about the compute budget.*