Before the conversation, Tomasz, co-executive director of the Ethereum Foundation: Why am I investing the next ten years in the robotics industry?

Source: The Round Trip

Compiled by: Yuliya, PANews

From founding the Ethereum client Nethermind, to serving as co-executive director at the Ethereum Foundation, and then leaving the crypto industry to start anew—Tomasz Stanczak’s career path almost mirrors the development of Ethereum itself.

However, in today’s era of the rapid rise of AI large models, he hasn’t continued to focus on blockchain. Instead, he has turned his attention to robots, automation, and the physical world. In his view, intelligence in the digital world is being quickly commoditized, while the real opportunities lie in the real world that still requires humans and machines to work together.

In the first episode of the 《The Round Trip》 xBuilders interview series, jointly produced by PANews and Web3.com Ventures, Tomasz looks back on his entrepreneurship and explains why he believes the robotics industry is positioned much like Ethereum was in 2016: still immature, but already on the eve of an explosion.

From finance to Nethermind, the Ethereum Foundation, and then robots

Host: Hello everyone, welcome to the latest episode of Round Trip. Today we have Tomasz Stanczak, who used to be the former co-executive director of the Ethereum Foundation and is also the founder of Nethermind. Tomasz, could you introduce yourself first? Over the past 10 years, how did you get involved in the blockchain wave?

Tomasz: I joined this industry in 2017. Back then, I always felt like I was coming too late—as if everything in the industry had already taken shape. So when people now call me an OG in blockchain, I actually find it a bit funny.

Before joining crypto, I worked in finance in London. I worked at banks and also spent time at hedge funds. Later, I started paying attention to the crypto industry and ultimately decided to leave the financial world. In 2017, I founded Nethermind. Over the years, we grew the company to a few hundred people, and our revenue remained quite stable. Then, in March last year, I joined the Ethereum Foundation and stepped down as CEO of Nethermind.

There’s an important reason for doing so: after joining the Foundation, I needed to eliminate all potential conflicts of interest as much as possible, which meant exiting many of my previous business relationships. After leaving the Foundation, it actually gave me more freedom. At Nethermind, I only kept my role as a board member and no longer handled day-to-day management. That finally let me think freely about what I want to do next.

After leaving the Foundation, the top three directions I most wanted to explore were: AI, robots, and the physical world.

Something important is happening in the digital world: intelligence is turning into a capability that is becoming cheaper and more widespread.

In the past, software itself was value. But with the development of AI, intelligence in the digital world is being rapidly commoditized. What’s truly full of friction and has huge room for efficiency gains is the real world: manufacturing, automation, logistics, data centers—there are a lot of problems waiting to be solved in these areas.

I originally only wanted to explore, but just a few days after resigning, I was invited to tour HF0’s residency program. There, I quickly got clarity on the direction. About a month later, the new company was formally registered and established.

Host: What does it feel like to start over again? After all, you’ve been away from hands-on engineering for a while now.

Tomasz: Entrepreneurship is constantly putting yourself into an “uncomfortable” environment. The HF0 residency was a fantastic experience, and I pushed myself further into “the pain mode”: I had to learn new things from scratch, or pick up rusty skills again. Over the past three or four years, I haven’t written a lot of code. But in the AI era, programming itself has changed a lot, which means I must build better systems-thinking architecture—knowing how to use AI Agents to construct systems.

A few weeks ago, I went to Detroit. There, manufacturing-focused entrepreneurs were building drone and robot and physical products. The whole city was full of energy. It reminded me of the early blockchain community: people met in person frequently, collaborated, and co-developed. Because robots belong to the physical world, they naturally need people to gather in labs, factories, or test sites. I believe many people who originally worked in blockchain will naturally move into these real-world “physical nodes.”

Robots are on the eve of a breakout; humanoid robots are just a transition

Host: We’ve heard about a “robot revolution” for years, but true large-scale applications never really showed up. Now a lot of capital is flowing into this space, and AI is rapidly driving the whole industry forward. Why is it precisely now that you’re willing to invest the future decade in this field?

Tomasz: The biggest change is that technologies that used to develop in a scattered way are now maturing at the same time.

First, AI models and training capabilities are advancing rapidly. At the same time, the entire robotics industry has been exploring different technical routes. Everyone is thinking about how to make robots truly useful.

AI isn’t only helping robots become smarter—it’s also changing robot R&D itself. For example, material design, mechanical structure design, and optimization of overall robot systems can now all be accelerated using AI.

In the past, every trial-and-error cycle in traditional manufacturing required a long timeline. Even changing a design might take six months to validate. Now, increasingly mature simulation environments let many experiments be done in the virtual world. Combined with falling actuator costs and progress in hardware supply chains, R&D speed has improved qualitatively. It’s like writing code with AI now: if you’re not satisfied, you just modify the prompt—much like how efficient that is.

People in the industry believe that this year is the most exciting year in robotics history. Even though many question whether “humanoid robots haven’t been deployed at scale,” the irreversible value of robots isn’t just “replacing humans,” but filling the labor gap in the real world. For example, the global data center industry currently lacks dozens of thousands of skilled maintenance technicians to handle maintenance and operations of data centers. Robotics companies are rushing into this market, hoping to fill the gap with automation.

So, in my view, this isn’t a “will robots appear or not” problem anymore—it’s a “which industries will robots enter first” problem.

Host: A lot of people are working on humanoid robots now—what do you think?

Tomasz: I think humanoid robots are only the beginning. At first, to enter the real world quickly, we’ll try to make robots look like humans, because then we can directly use existing human work environments and complete tasks more easily via remote operation. But over time, we’ll gradually realize that not all jobs are suited to a human form.

In the future, we’ll design robots in different forms for different tasks.

Some machines may not need two legs or two hands. They’ll become machines optimized for a specific job, not just a copy of human appearance.

This trend will show up across industries.

Agriculture needs different robots. Healthcare needs different robots. Manufacturing needs different robots.

In fact, logistics and warehousing have already proven this. Today, large warehouses are highly automated, and in the future, more and more industries will go through the same process.

Today’s robots remind me of Ethereum in 2016

Tomasz: It feels a bit like how I felt about blockchain when I saw it in 2017. In 2016 and 2017, Ethereum started to receive attention. Despite criticisms at the time—slow speed, lack of maturity, and security issues—people could already see a clear development roadmap, and they gradually believed the technology could solve real-world problems. Today, stablecoins have become one of the most successful applications of blockchain, and over the past two years, their usage scale and adoption curve have grown quickly.

Also, I think people were initially excited about “AI Agent payments on blockchain,” then the hype cooled for a while. But I expect it will regain some attention, and I’ve seen some interesting data.

The robotics industry is going through a similar process. Even if it isn’t mature enough yet and many products are still in the lab stage and haven’t been deployed at scale, people are starting to see the potential for robots to perform complex tasks. As more and more people believe this technical route is viable, the industry’s development speed will accelerate rapidly.

In fact, even when I founded Nethermind, I had long been interested in autonomous machines and robots. Many people may not know that the company name behind Nethermind (Demerzel Solutions Limited) takes “Demerzel” directly from the core character of Isaac Asimov’s famous science fiction works: the humanoid robot Eto Demerzel from the Foundation series and the Robot series.

And my new company is continuing this tradition, paying tribute to Asimov as well. For me, this isn’t a sudden pivot—it was a seed planted a decade ago, and only today did it finally reach the moment when it truly matured.

*Note: Isaac Asimov is known as the “founder of robotics,” a well-known American science fiction writer.

Blockchain payments will ultimately belong to AI—not humans

Host: Many people used to think that the real application scenario for crypto isn’t payments between people, but payments between machines. Not long ago, I was chatting with a founder about this. He said humans have already “trained” AI into a kind of consensus: traditional payment experiences are poor—there are chargebacks, and it can take days for funds to arrive. In contrast, crypto payments are almost like they were prepared for AI.

So I’m curious: when you restart now, will you also bring your accumulated experience from the blockchain space into the robotics industry? Or, in your view, how many years will it still take for real machine-to-machine payments to arrive?

Tomasz: Right now, I’m not focusing on blockchain or crypto-related parts in robots. I’ve fully turned to automation in the physical world—especially automation deployment (like operating systems and autonomous business deployments). After only a month and a half, many things are still being defined as we refine the vision. I don’t need to think about payments right now.

Of course, the company has only been around for one or two months, so many ideas are still at the stage of constant improvement. At this point, we’re more defining our vision than rushing to announce a final product. But as the research goes deeper, the direction is getting clearer and clearer.

Still, in the future, I believe blockchain payments being used by Agent AI and robots is a very natural thing.

I’ve always argued that blockchain provides the underlying framework: it should be invisible and natural for both AI and humans. It’s simply a technology that makes everything more reliable and lowers the cost of trust.

If an infrastructure layer is good enough, it should be invisible by nature.

In interviews, I’m often asked about “buying coffee with blockchain and user experience.” My answer is often seen as “heretical.” I don’t believe blockchain is directly for users. Human beings never directly verify blockchain (except for a few extremely rigorous people who run their own nodes and understand the mathematics behind cryptography). Most of the time, there are tools between you and blockchain, like block explorers or wallets. If those tools are centralized, that becomes the weakest link and it can easily show you forged information. A recent announcement from the Ethereum Foundation also emphasized this access-layer problem.

I strongly believe that AI’s role between users and blockchain is crucial. It helps users receive and verify information, because it’s often hard to verify information in the real world. In the long run, we need “verifiable sensors” and “verifiable AI” to reveal the real world. Humans don’t always pursue “truth” in daily life—for example, when watching TV dramas, we accept some content that isn’t fully real, but occasionally we still want to participate in genuine reality. In such cases, relying on AI, verification mechanisms, blockchain, and verifiable sensors is the only solution, even though implementing this in the physical world remains challenging.

Regarding whether robots will use blockchain payments and chargeback mechanisms, I think chargeback mechanisms will reappear on-chain. The underlying transactions need to remain in a successful or failed state, while the upper layers can build chargeback mechanisms to verify information and reduce trust assumptions. Such a mechanism may help robots find a balance between trust and latency, enabling the system to act faster and provide additional protection.

Also, when robots choose a blockchain, they may rely on Ethereum, Layer 1, Layer 2, or a native chain, but ultimately it will be controlled by some AI Agent. Even if we think robots will be localized, in reality they’re often a group system managed by an Agent controlling multiple edge devices or machines. This integration will provide a more efficient operating approach for future intelligent systems.

A real-world version of Factorio: automatically deploying data centers

Host: What exactly do you mean by “deployment automation” you just mentioned? Since you’re still in stealth mode, can you share a bit more about what you’re building?

Tomasz: Of course. It’s not that we’re intentionally stealth—it’s just that we haven’t spent time making marketing materials yet. During our time at HF0, I’ve been looking for that “zero point” to focus on.

What we’re doing right now is “data center deployment.” It’s actually highly related to blockchain’s “decentralized node” concept. A lot of people now complain that running Ethereum nodes requires too many resources and can only be done in the cloud or in data centers. True decentralization should mean you can deploy a node extremely quickly, along with its own data center, so it can quickly connect to physical space.

The role of robots here is to make these data centers operate and become autonomous. We’re not trying to replace humans—we’re using machines to solve labor shortages and optimize the entire deployment workflow. It’s not just logistics and supply chains, but also legal approvals, permissions, and so on. We’re optimizing everything related to “building in physical space,” whether it’s energy, data centers, or drone swarms.

When AI becomes our customer, it may need to generate its own source of intelligence in different spaces. But to AI, human interfaces and deployment workflows are too slow and too frustrating. It’s like how we used to think cross-cloud server migration was torture, which led to the birth of GitHub and various cloud SaaS solutions. What I want to build is a solution from “GitHub to physical space.” I’ve mentioned the concept of “social GitHub” before when speaking at Devcon in Cannes and Bangkok: you just need to press a button, provide funding, and it deploys itself.

We often talk about building Factorio in the real world. In this game, everything starts from the underlying compute capability and energy. Yes, you could call it “recursive deployment”: an Agent deploys robots, the robots deploy other robots, helping you self-build faster. It sounds like science fiction, but the accelerating progress in AI, materials, and automation is making this kind of self-reinforcing loop possible.

*Note: Factorio is a game about building and maintaining factories. In the game, players need to mine resources, research technologies, build infrastructure, perform automated production, and fight off alien enemies.

Turbulent Ethereum years: transparency, delivery goals, and decentralized governance

Host: It sounds like the real breakthrough is giving intelligence the ability to experiment and build in the physical world. Speaking of that, I’m also curious about your experience at the Ethereum Foundation. You witnessed the construction of the most important technology infrastructure up close. Now that we step back a bit—when you joined, what did you think the Foundation was doing wrong?

Tomasz: When I joined, I indeed went in with the goal of driving organizational structure changes, and everyone was ready to accept these difficult changes. We wanted the core organization to become more hierarchical and more goal-oriented, while also being more transparent.

The biggest problem at the time was that the Ethereum Foundation wasn’t brave enough in communication. This mainly came from the unusually strict regulatory environment in the early years. Fortunately, the environment was starting to improve before I joined, so my task was relatively easier—I only needed to tell everyone, “Now we can communicate boldly.”

Back then, many infrastructure operators—such as L2 operators, DVT, staking projects, Consensys, Bitmine, Nethermind, and others—started speaking up for Ethereum actively in public. That also meant the Foundation was no longer the only node that spoke. We began to clearly answer what we were doing and why we were doing it, refusing the mindset of “we made the decision, but we must do it secretly.” Internal communication and external communication were equally important. If you weren’t transparent externally, you’d also worry about leaks internally and stop communicating. We solved this by activating the Ethereum official Twitter account and starting frequent conversations with big institutions, including enterprises and banks.

In terms of management, the core goal I set was: keep delivering fork upgrades twice per year. This precisely defined the role of the Ethereum Foundation. It shouldn’t tell everyone “what must be developed.” Instead, it should act as a coordinator—providing high-quality testing, security reviews, research guidance, and funding support—to ensure that core developers can deliver on time. As long as you make “on-time delivery” the goal, you can reorganize all internal accountability mechanisms around it. Previously, the Foundation might have been too focused on absolute decentralization, believing that hierarchical structures should be excluded as a centralizing force. But some degree of goal-orientation is necessary.

Host: Recently there have been news that the Ethereum Foundation is shrinking its scale, while another nonprofit, Etherealize, says it will take on more R&D work. That sounds like good news—what do you think?

Tomasz: After I left, there indeed was a noticeable shift: the direction became less focused on the enterprise side and more on privacy. I think that’s actually good for the Foundation. It really needs to return to these core roles.

However, I don’t like the way this kind of shift is implemented—such as requiring someone to sign a forced agreement just to stay. The kind of “Milady”-style culture in it isn’t my vibe. It makes me feel like someone is trying to define Ethereum into a specific shape or style. That’s completely the opposite of what I hope to see: I want Ethereum to remain absolutely neutral, allowing different paths and giving everyone equal opportunities to participate in building.

But I fully agree that the Foundation needs to regularly steer back to its core mission—values, the cypherpunk spirit, anti-censorship, open source, privacy, and security. If the Foundation realizes it’s just one node among many, it can narrow and focus its responsibilities.

On the other hand, the newly formed EthLabs brought together a group of highly talented leaders who are more focused on deploying and adopting Ethereum in areas like finance, and even AI. People who stayed and those who left to build the new team are all excellent. Even though there are differences in direction, I think these healthy technical debates are very valuable. Developers love to argue, and I’m looking forward to seeing both sides contribute great things.

The “private chain” narrative is being eliminated

Host: Since we’ve talked about deployment use cases—Nethermind recently announced a partnership with UBS to explore privacy technology on Ethereum. I believe you must have been looking forward to this day eight years ago.

Tomasz: Yes, that’s very reassuring. It also demonstrates the outstanding capabilities of the Nethermind management team, led currently by Daniel Salazar. These talents have fought alongside us for years, are extremely loyal to our vision, and have achieved great success in the enterprise market.

We made an early judgment that 2025 and 2026 are key time nodes for deep dialogue with enterprises. As the regulatory environment thaws, stablecoins are widely adopted, and the rise of asset tokenization emerges, innovation teams inside banks that were previously holding back start truly building. Nethermind seized this opportunity—leveraging long-term accumulated engineering and research strength while collaborating with top global financial teams.

Enterprise-grade privacy is a very difficult problem—protecting user privacy while also meeting compliance requirements, and preventing blockchains from being misused once they become usable. As far as I know, Etherealize is also focusing on enterprise privacy and asset-category tokenization.

When large financial institutions finally decide to choose public chains, that’s the biggest confirmation of the publicity strategy we promoted last year. (We kept calling for “build it on public chains—doing experiments on private chains no longer makes sense.”) Now, early non-public-chain projects like R3 and so-called “alternative chains” (many market themselves as public chains, but in reality are similar to private chains) are being eliminated by the market, while Ethereum implements truly permissionless in a very rigorous way. Seeing them ultimately return to public chains is a better outcome for everyone, because it means higher security.

Eliminate latency and build an efficient coordination layer

Host: This conversation has been fascinating. From Ethereum to AI to robots—these things at seemingly different levels ultimately come down to the problem of “coordination.” If you put everything you’ve done—finance, Ethereum, and robots—together, what are you ultimately trying to build?

Tomasz: For me, I’ve been doing the same thing: building a coordination layer, optimizing the coordination process, and eliminating latency from all systems. That’s the core element running through all my work.

For example, at Nethermind, “the fastest sync time” is always one of our ultimate goals. In the past, every morning, the core engineering team tracked our node market share. When we reached 33% market share, the team would proactively help other client teams in the ecosystem (the mutual-help culture among Ethereum clients is great). But beyond that, we were obsessed with performance optimization: how fast can we sync the mainnet under different modes? How fast can we sync different chains?

The same logic applies at the infrastructure level: how fast does it take to deploy a node? How fast does it take to sync to the latest blocks? These are all about latency—how quickly you can reach the state where you can proactively participate in block-building decisions?

This is completely transferable to what I’m doing in the physical world now. How fast can you deploy in physical space a data center that lets nodes run? This kind of optimization on the physical layer ultimately boils down to identifying bottlenecks and continuously judging which bottleneck is most important to solve at the moment. This not only defined my past, but also defines the work I’m fighting for now.