RLWRLD Robot AI: Korea Captures Craftsmen Skills

Samantha: Welcome to The State of Tech, The European Edition, Wednesday May thirteen, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: a South Korean startup capturing human expertise to build universal AI brains for robots, a proposal to tax AI profits and redistribute them across society, Europe's JUPITER supercomputer smashing a quantum simulation record, Appian rolling out agentic automation for enterprises, the UN pushing AI for the future of cities, and Unitree unveiling a manned mecha you can actually buy. Let's start with those robot brains.

South Korean startup RLWRLD builds universal AI brains by recording how master craftsmen actually use their hands.

Samantha: We're opening in Seoul. A startup called RLWRLD, pronounced "real world," is building what they call universal AI brains for robots. And their method is fascinating. They're not just scraping the internet. They're going into hotels, warehouses, and convenience stores to capture how skilled humans actually do things with their hands.

Bob: The partner list tells the story. Lotte Hotel is contributing food and beverage techniques, so think plating, pouring, knife work. CJ is handing over logistics know-how. And Lawson, the convenience chain, is showing how staff organise retail displays. All of that flows into a robotics foundation model RLWRLD just unveiled.

Samantha: The target is broad deployment in industrial sites and homes by 2028. Ambitious, but they're betting human-shaped machines are the future, and to make those work you need human-level dexterity, not just human-level chat.

Bob: And there's a national layer here. The South Korean government is funding projects to database what they call master technicians' know-how. The country has an aging workforce, productivity is a real concern, so capturing expert skill before those experts retire is treated as strategic infrastructure.

Samantha: Which is the part Europe should be watching closely. We have the same demographic pressure. Germany, Italy, Spain all face shrinking pools of skilled trades. If Korea builds a library of how a master welder, chef, or warehouse picker moves, that becomes exportable IP.

Bob: And it raises a question European policymakers haven't really tackled yet. Who owns the data of a craftsman's hands? If a Michelin-trained chef trains a robot, is that the chef's IP, the restaurant's, or the AI company's? Korea is moving while we're still drafting frameworks.

Samantha: The practical takeaway is that the next wave of robotics may not be won by whoever has the biggest model. It'll be won by whoever has the richest library of physical human skill. Data of the hands, not just data of the web.

Bob: For European industrial groups, Siemens, ABB, KUKA, the message is clear: partner with skilled trades now, or import the brains later.

A senior Korean official proposes taxing AI excess profits to fund basic income and startups.

Bob: Still in Korea, but a very different angle. Senior Presidential Secretary for Policy Kim Yong-beom has floated what he calls a national dividend. The idea: tax excess profits generated by AI and the semiconductor boom, and redistribute that revenue across society.

Samantha: He's pointing directly at Samsung Electronics and SK Hynix. Both are riding the AI chip wave, and the worry is that this concentrates wealth so heavily it deepens social polarization. The proposed uses include startup support for young people, basic income for rural communities, and stronger pensions.

Bob: What's interesting is how he frames it. Not as redistribution, not as a wealth tax in the traditional sense, but as a system maintenance cost for the AI era. The argument being: if you want social stability while you transition to a tech-dominated economy, you have to fund the transition.

Samantha: That framing matters because it sidesteps the usual political fight. Calling it maintenance rather than redistribution is clever positioning, and European finance ministers will notice.

Bob: And the European parallel is obvious. The EU AI Act covers risk and rights, but it doesn't address the windfall question. If Nvidia, ASML, and a handful of hyperscalers capture most of the AI economic upside, what happens to everyone else?

Samantha: France and Spain have flirted with digital services taxes before. This Korean proposal could revive that conversation with a sharper edge, because it's specifically tied to AI profits rather than digital revenue broadly.

Bob: There's also a competitive risk. Tax AI too aggressively in Europe and capital flows elsewhere. Tax it too lightly and you get the inequality Kim is warning about. Nobody has cracked that balance yet.

Samantha: It's still just a proposal in Seoul, not law. But the language of national dividend tied to AI profits is new, and once a phrase like that enters the policy bloodstream, it tends to travel.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

Europe's JUPITER supercomputer simulates a fifty-qubit quantum machine, breaking the previous world record.

Bob: Big European win here. Scientists at the Jülich Supercomputing Centre, working with NVIDIA, have simulated a universal quantum computer with fifty qubits. That's a new world record, beating the previous mark of forty-eight qubits, also set by Jülich back in 2019.

Samantha: The machine doing the work is JUPITER, Europe's first exascale supercomputer, which only went live last September. It's running NVIDIA GH200 Superchips, and this simulation is essentially the first major demonstration of what an exascale European system can do.

Bob: Each extra qubit roughly doubles the complexity. So going from forty-eight to fifty isn't a small step, it's four times harder. And simulation matters because real quantum hardware at this scale is still unstable. Simulating lets researchers test algorithms before the physical machines catch up.

Samantha: Which is the strategic point for Europe. We've talked a lot about lagging behind the US in AI chips and behind China in scaling. Quantum is one area where Europe still has genuine leadership, through groups like Jülich, IQM in Finland, and Pasqal in France.

Bob: JUPITER being operational gives those teams a home-grown platform to run on. They don't have to rent cycles from American clouds. That's sovereignty in the practical sense, not just the political slogan sense.

Samantha: And the algorithms developed here have real downstream uses, drug discovery, materials science, cryptography. The pharma industry in particular, Bayer, Sanofi, Novartis, is watching quantum simulation closely for molecular modelling.

Bob: One caveat. Simulation is not the same as a working quantum computer. We're still years away from fault-tolerant machines that can do useful work at scale. But fifty qubits simulated on European silicon is a credible milestone, and it puts Jülich back on top of the leaderboard.

Appian adds spec-driven AI development and agentic automation to its enterprise platform.

Samantha: Enterprise software news. Appian has rolled out a significant update integrating AI-assisted spec-driven development and something called Model Context Protocol, or MCP, for agents. The goal is to embed AI directly inside business processes rather than bolting it on the side.

Bob: The pitch addresses two real problems enterprises keep hitting. Fragmented data, and lack of control over what AI agents actually do. Appian's answer is structure, context, and guardrails, so agents stay smarter, safer, and auditable.

Samantha: The new capabilities let agents interface securely with external enterprise systems. They get access to Appian's data fabric, which gives unified data access across silos. And crucially, organisations can track agent performance, apply learned memory across processes, and optimise objectives without losing oversight.

Bob: Model Context Protocol is the interesting bit technically. It's an emerging standard for how agents talk to tools and data sources. Anthropic pushed it originally, and now we're seeing platforms like Appian adopt it. That's important because without standards, every enterprise ends up with bespoke agent plumbing that doesn't talk to anything else.

Samantha: For European enterprises this matters because of compliance. GDPR, the AI Act, sector-specific rules in finance and healthcare, they all require traceability. An AI agent that just acts without an audit trail is a regulatory nightmare. Appian is selling the audit trail as a feature.

Bob: It also lowers the barrier for less technical teams. Spec-driven development means you describe what you want in structured terms and the platform helps build it. That's faster than traditional coding and more controlled than just prompting a model.

Samantha: The competitive picture is crowded. ServiceNow, Salesforce, Microsoft Power Platform are all chasing the same agentic enterprise space. Appian's bet is that governance and process integration matter more than raw model power, and for European buyers worried about compliance, that argument lands well.

Bob: Worth watching whether MCP genuinely becomes the cross-vendor standard, or whether it fragments into competing flavours.

The UN urges cities to embrace an AI-enabled citiverse for smarter, more resilient urban living.

Bob: To close out, two things you can actually use today, sort of. First, the UN Virtual Worlds Day in Geneva. Twenty UN organisations and urban experts gathered to push governments and the private sector to harness AI and digital innovation for the future of cities.

Samantha: The framing is the AI-enabled citiverse. That's AI combined with spatial intelligence, digital twins, and immersive tech, all stitched into urban planning. By 2050, seventy percent of the world will live in cities, so the stakes are concrete.

Bob: They issued a Call to Action for Humanity with five strategic priorities. Building trusted AI systems, strengthening international cooperation, and ensuring AI contributes to efficient, resilient, responsive communities are the headline ones.

Samantha: For the listener, here's the practical hook. Digital twin platforms for cities are increasingly open. Helsinki has a free public digital twin you can explore in your browser. Rotterdam, Vienna, and Amsterdam all have open city data portals with three-D models you can download.

Bob: If you've ever wondered how your neighbourhood is being planned, or you want to see proposed building heights, traffic flows, or solar potential on your roof, these tools are genuinely accessible. No special software. Open the website, click around.

Samantha: Helsinki's is probably the most polished. Search Helsinki 3D plus. Free, runs in the browser, works on a laptop in about five minutes. Vienna's smart city portal is in German but translates cleanly. Amsterdam's 3D viewer shows planned construction citywide.

Bob: It's a nice antidote to abstract AI hype. You can literally see what AI-enabled urban planning produces in your own city, today, without installing anything.

Samantha: And if you're a homeowner, the solar potential layers in some of these tools are surprisingly useful. Click your roof, get an estimated annual yield. Free, fast, no signup.

Bob: Worth half an hour of your afternoon.

Unitree launches the GD01, a production-ready manned mecha you can actually buy for six hundred fifty thousand dollars.

Samantha: And the second thing, a bit more aspirational. Unitree Robotics in China has unveiled the GD01, billed as the world's first production-ready manned mecha. Roughly two point seven metres tall, a human pilot climbs inside and drives it.

Bob: The launch video shows it smashing through walls with mechanical arms. Pricing starts at three point nine million yuan, about six hundred fifty thousand dollars. So not exactly an impulse buy, but it is on sale.

Samantha: The Chinese government's been pushing humanoid robots and embodied intelligence hard. February this year, the Ministry of Industry and Information Technology released a standard system. Beijing, Shenzhen, and Shanghai have all rolled out supportive policies and funding.

Bob: For most of us this isn't something we're buying. But Unitree's smaller quadruped robots, the Go2 dog, are actually consumer-priced, starting around fifteen hundred euros, and ship to Europe. If you want to play with serious robotics at home, that's the realistic entry point. Programmable, app-controlled, surprisingly capable.

Samantha: The GD01 is more of a statement than a product. It signals that Chinese robotics is willing to ship hardware Western firms are still calling research projects. For European manufacturers, that's the competitive reality.

Bob: Whether mecha pilots become a job category, we'll leave to science fiction. But the gap between concept video and product page just narrowed considerably.

Samantha: Today we covered: Korean robot brains learning from master craftsmen, a proposal to tax AI profits as a national dividend, JUPITER simulating fifty qubits for Europe, Appian's agentic automation for enterprise, the UN's call for AI-enabled cities, and Unitree's manned mecha hitting the market.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.