Skip to content

The CCS Interview

Cartesi's Argento: stage two rollups let any honest user enforce the rules

Felipe Argento, co-founder of Cartesi, explains how the project's Linux-based virtual machine brings Web2 developers and tools to Ethereum. He also covers what a stage two rollup means, why Cartesi is turning toward DeFi, and the plan to make stage two chains easy to launch. The conversation was recorded in January 2026.

Interview by Ashton Addison Recorded January 2026 6 min read

Key takeaways

  • Cartesi is a Linux-based virtual machine layer 2 that enables developers to use traditional programming languages and existing open-source libraries on Ethereum.
  • App-specific rollups built on Cartesi inherit Ethereum's security and decentralization while allowing different computational capabilities than EVM-based layer 2s.
  • As Ethereum scaling improves, layer 2s will need to differentiate beyond speed; Cartesi's competitive advantage lies in developer accessibility and compatibility with web two ecosystems.
  • Millions of existing open-source libraries and future code written for Linux systems become available to blockchain applications through Cartesi's architecture.
  • AI coding agents are expected to be more effective at building applications compatible with traditional computing environments than pure EVM-based web three applications.

A Linux machine on Ethereum

Ashton Addison: What is Cartesi, and where does it fit in Ethereum's scaling picture?▶ 01:43

Felipe Argento: I like to explain Cartesi differently every time, so let's go with this: Cartesi is an addition to Ethereum. We're adding a powerful engine on top of it that does things Ethereum normally does not do. It inherits Ethereum's properties: decentralization, censorship resistance, reduced counterparty risk, everything institutions like. The new power is mainly a new machine. We call it the Cartesi Machine, a Linux-based virtual machine that extends what apps can do without losing the things that make Ethereum amazing. The technical term is an app-chain rollup with a different virtual machine from the EVM, one that lets you do things you couldn't do with Solidity. And it looks like a normal, traditional computer. You can use the programming languages and libraries you'd use for Web2 applications, and since we have an operating system, you can do virtually everything you can do on an operating system.

Ashton: Why does a Linux-based VM matter for developers?▶ 04:08

Felipe: You have the number of developers who know how to code for Linux, plus all the open-source code that has been written in the past. We've run TensorFlow, an AI library, on the Cartesi Machine, so it's now compatible with the blockchain. Even more incredible, it also covers the future code being written. Web2 hasn't come to a halt and fully migrated to Web3. The majority of development work is still on Web2. So the bridge we built between traditional machines and the Ethereum Virtual Machine also makes all future open-source code available to blockchain applications.

Ashton: Does AI change the picture for developers?▶ 08:41

Felipe: AIs learn by consuming lots of data. Coding agents learn to code by looking at existing code, and there's a lot more code for these normal machines than for the EVM. So I expect coding agents to help much more with building an app that's compatible with Web2 than with building a pure Web3 app. That's not a good way to put it, though, because apps built on Cartesi are pure Web3 apps. They're unstoppable, they run in a decentralized manner, and they're censorship resistant. They just use the same languages that Web2 apps use.

You cannot stop it from running and you cannot mess with the rules the code has defined, just like Ethereum.

Standing out among layer 2s

Ashton: How is Cartesi different from layer 2s like Base and Arbitrum?▶ 05:46

Felipe: It's similar in some ways and very different in others. Historically, layer 2s didn't try to differentiate from Ethereum. They just tried to be more scalable. In a sense they're a copy of Ethereum with bigger block sizes, or sequencers to reduce latency. That's a fair position, but I don't think it's the best one. Ethereum is now focusing on scaling the L1 again, so using Ethereum mainnet is going to get much cheaper. What happens to all these EVM L2s that copy Ethereum's functionality but are faster? Cartesi is similar in how a rollup is constructed: the inputs come from the blockchain, and every state is verifiable on Ethereum and enforced by Ethereum. The significant difference is the machine. Base is an EVM layer 2, so it runs EVM code. We run a Linux machine. I think we're going to go through a phase where these L2s are going to be forced to differentiate, since scaling is no longer the bottleneck.

Ashton: Does Cartesi need a marketing push to convince Web2 developers, or does the technology speak for itself?▶ 11:12

Felipe: That's a current challenge. When we began in 2018, we went to Web2 developers because we allow Web2 code. But the sales pitch meant first convincing them to use Web3, then explaining Ethereum, then explaining that Ethereum is incomplete because of a scalability problem, then explaining rollups. By the time we got to Cartesi, they weren't interested anymore. That didn't work. Now that institutions are coming to Ethereum, the conversation is much easier. They know what Ethereum is, and they know it can be an upgrade to the traditional financial system. So we can say: if you want to be involved, you can use the coders your bank already employs, have them use Cartesi, and you're suddenly plugged into this new world.

Ashton: Does being an app-specific rollup help with institutional adoption?▶ 13:54

Felipe: Yes. Another difference from Base or Optimism is that Cartesi is an app-specific rollup. For each application you spin up a new chain, and all these chains have the characteristics we've discussed. That gives you a lot of power. For institutions it's much better, because they can spin up a chain that is compliant with whatever regulations they have to comply with, and with the characteristics they need. I think we're in quite a good place to help onboard institutions.

Stage two and trust

Ashton: Does the Cartesi Machine still rely on Web3-style decentralized security?▶ 09:49

Felipe: Yes. The Cartesi Machine runs on top of Ethereum and inherits its trust assumptions, with one added assumption: there's one honest party using the application. If you have a single honest person using an application running on Cartesi, you have all of Ethereum's assumptions: decentralization, censorship resistance, resilience. The cool thing is that the honest person can be you. You're not relying on a fixed set of people, and you can enforce whatever you're using on Ethereum yourself. That's non-negotiable. Otherwise, you just run whatever you want on Amazon and then just trust the result.

Ashton: Cartesi is described as a stage two rollup. What does that mean, and why does it matter?▶ 15:51

Felipe: Ethereum released a rollup-centric roadmap. In theory, a rollup extends Ethereum's trust assumptions: it's censorship resistant, secure and resilient. In reality, many teams called themselves rollups before they had a proof system, and some were just running centralized servers. The landscape became a mess. So an organization in the space came up with a standard based on stages. Stage zero rollups still have a lot of centralized pieces. If your money is there, some people have the power to upgrade the system and could steal your money somehow. Stage one has a working proof system, but there's still some oversight: people can step in to fix bugs or upgrade the system. Stage two is totally decentralized and permissionless. It runs with the same rules as Ethereum, and no one can stop you from removing your money. Even if the founder or developer wanted the application to act dishonestly, you as a user can enforce honest behavior. Cartesi managed to accomplish that, and you can only do so if you take research and development seriously. There's pressure in this space to prioritize speed and convenience over robustness, because robustness only matters when things go wrong. We did the work. We created a proof system with a fraud proof game, everything implemented and running, and people can try to break it. We want to create a factory of stage two rollups, so you can build your application on Linux and deploy your Cartesi chain. As soon as you deploy it, it's no longer yours. You cannot stop it from running and you cannot mess with the rules the code has defined, just like Ethereum.

I think we're going to go through a phase where these L2s are going to be forced to differentiate.

DeFi focus and 2026 plans

Ashton: How important is DeFi on Cartesi right now?▶ 20:56

Felipe: That's an adjustment Cartesi is making. In the past we wanted people to create all sorts of things, so we saw experiments using AI, some very cool games and governance systems. We were exploring the world-computer side of Ethereum rather than its hyper-financialization, especially because DeFi was just beginning and quite simple. But even Vitalik said Ethereum is for finance and a long tail that includes governance and games, and the clear winner so far is finance. I don't quite like the TVL metric. Who celebrates money locked? You should celebrate money moving around and things being used. So we're coming to terms with the fact that the moment is mainly for DeFi. We're asking what kinds of DeFi applications need the computational power and expressiveness of the Cartesi Machine. How much more complex should DeFi be able to be? What financial libraries do people use on Web2 that could be used on Web3? How can you design a better order book with the tools we have for Linux?

Ashton: What are Cartesi's main goals for the first half of 2026?▶ 25:36

Felipe: Last year we achieved stage two. Now we're very close to generalizing it, so anyone can very easily launch a Cartesi chain that's already at stage two. I think that's going to be super cool. Then we have to work on onboarding developers and users, and maybe build something ourselves to show the power of Cartesi.

This interview has been edited for length and clarity. Recorded January 2026.

Felipe Argento
Felipe ArgentoFelipe Argento is a Brazilian blockchain engineer and co-founder of Cartesi, and he serves as an advisor to the Cartesi Foundation. He joined Cartesi in its early stages and developed the first version of its RISC-V Solidity step, an implementation of the full RISC-V instruction… Full profile →
Share this interviewXLinkedIn
▶ Watching the interview