What Is Cartesi (CTSI)? Linux Rollups, PRT Fraud Proofs and the October 2026 Shift

Cartesi is open-source infrastructure that gives each application its own rollup. The application's code runs off-chain inside a deterministic RISC-V machine that boots Linux, and the results are verified and settled on Ethereum. CTSI is the project's ERC-20 token. Holders can stake it through pools that produce blocks and earn rewards. The newest development came on 1 October 2026. The Cartesi Foundation said it is moving more of its engineering and business-development effort to an independent derivatives venue built on Cartesi Rollups. It funded that venue with a grant and holds a contractual right to 25% of its net revenues.

Most of this article relies on Cartesi's own documentation and blog. The main independent check is L2BEAT's listing of the project's Honeypot deployment.

The latest: the Foundation's October 2026 change of focus

In "The Next Stage for Cartesi" (1 October 2026), the Cartesi Foundation wrote that "technical advantage alone had less influence on adoption than we expected." It said established networks with users, liquidity and distribution counted for more. Its answer is to support a commercial product that runs on the stack under real demand:

• What changes. The Foundation is "directing more of its engineering and business development efforts toward supporting" a derivatives venue built on Cartesi Rollups. Its September 2026 update says the venue is "bringing options onchain." The venue runs through a separate corporate structure, independently of the Foundation, and its own team controls the product, launch timing and commercial projections.
• Who runs it. The founders named are Felipe Argento, Danilo Tuler, Pedro Argento, Eduardo Barthel and Brandon Isaacson, all from the Cartesi core team. The June 2026 update (published 3 July 2026) had already reported that Felipe Argento and Brandon Isaacson left the Foundation board "to start building on Cartesi."
• The money link. The Foundation funded early development through a grant. It holds a contractual right to 25% of the venue's net revenues, after essential operating and growth costs, and says it will use that revenue for Cartesi development.
• What stays. Cartesi remains open source and general purpose. The Foundation names its infrastructure priorities as completing the integrated Rollups and fraud-proof stack, with security testing and hardening, reliability, verification and fund recovery at the top. It also says it has discontinued several parallel initiatives.
• CTSI. The token itself does not change. The post says "any future evolution of CTSI's utility or economics would be considered separately" under the applicable governance process.

How Cartesi works

The Cartesi Machine

The documentation calls the Cartesi Machine "Cartesi's solution for verifiable computation." It is an emulator of Cartesi's own flavour of the RISC-V architecture, with a 64-bit physical address space. It runs a Linux kernel with an embedded Linux distribution. The machine is deterministic: "Two parties performing the same computation always obtain exactly the same results." This lets heavy computation run off-chain while the chain only steps in when results are disputed.

Cartesi Rollups

Cartesi Rollups are "application-specific, assigning each dApp its rollup app chain and CPU." The flow works like this:

1. Inputs start on-chain, on the base layer.
2. A Cartesi Node reads them from the data-availability layer.
3. The machine processes each input as an advance_state request.
4. The app emits outputs. Notices are provable outputs that can be verified on an EVM chain.
5. Once the optimistic-rollup dispute window passes, outputs become "verifiable and possibly executable on the settlement layer."

Because the machine runs Linux, developers can use "their preferred programming language or tool" rather than a contract-only language. The docs' overview example is written in Node.js.

PRT: the fraud-proof system

PRT (Permissionless Refereed Tournament) is the core algorithm of Cartesi's fraud-proof system:

• Claims. Each claim commits to the whole computation path as a Merkle tree over every state transition, not just the final state.
• Tournaments. Conflicting claims meet in bracket-style matches. Each round eliminates half of them, and the last disagreement is settled by a one-step proof on-chain.
• Sybil resistance. Every competing claim must post a bond, while the honest side needs only a single stake. The honest party's workload grows logarithmically with the number of attackers.
• Security assumption. The system is designed so that 1-of-N honesty is enough, and the docs say "anyone - even on a modest computer - can be a participating validator."

Independent check: the PRT Honeypot on L2BEAT

Cartesi runs a "Honeypot" on Ethereum mainnet: a token vault secured by PRT that anyone can try to break. L2BEAT lists Cartesi PRT Honeypot v2 as a "Stage 2 Appchain Optimistic Rollup." Its listing reports fraud proofs with a "7d 1h challenge period," data availability on-chain, self-sequencing and self-proposing.

L2BEAT also flags the limits. "The single hardcoded address can withdraw all funds," and users cannot exit because deposits are treated as donations. "The challenge protocol can be subject to delay attacks." When checked on 25 October 2026, the value secured was about $1.40K, all of it in CTSI. Cartesi's own report of the first Honeypot says it surfaced a liveness bug, and the v2 deployment followed.

Engineering status as of Q3 2026

Cartesi's Engineering Report for Q3 2026 (1 July to 30 September, published 15 October 2026) says the stack is still in pre-release:

• Cartesi Machine Emulator v0.21.0 shipped on 4 August 2026, with matching guest-tools v0.18.0, Linux image v0.21.0 and machine-solidity-step v0.15.0 releases.
• Deposit refunds. Rollups Contracts 3.0.0-alpha.7 (merged 22 July) lets anyone refund unfinalised Ether, ERC-20, ERC-721 and ERC-1155 deposits to their depositors once an application is foreclosed. Node support was still in progress at quarter end.
• Emergency withdrawals for finalised deposits use an account-proof path delivered in Q2. Cartesi's next milestone is end-to-end availability.
• Claim staging (merged 15 July) made settlement two-phase. "Sentries can speed up acceptance but cannot block or change a result."
• First PRT validator. The Sling node (merged 27 July) is described as the "first operational PRT validator," now "in testing and hardening."
• Still alpha. The fraud-proof system is at v3.0.0-alpha.5. Rollups Contracts v3 reached alpha.10 on 31 August and is deployed on the Ethereum, Arbitrum, Base and OP Sepolia testnets. The report lists no new mainnet deployments. PRT targets Ethereum today, and other chains are experimental.
• In the browser. The September update introduced @cartesi/machine 1.0.0-alpha.1, which runs the Cartesi Machine API in web apps through WebAssembly.

The CTSI token

Staking

According to Cartesi's staking documentation:

• You delegate CTSI to a pool through the Staking Portal. You first approve an allowance, then deposit and stake.
• "For every block the pool produces you will get a share of the reward minus the commission taken by the pool." Rewards compound automatically.
• There is no minimum ("you could even stake only 1 CTSI"). The docs warn that very small stakes may take a long time to produce blocks.
• Block production is probabilistic.
• Withdrawals are possible after an unstaking lock period, but the FAQ does not state its length.
• Each pool sets its own commission.

The staking FAQ also quotes a per-block reward figure. The page does not date that figure, so it is left out here. Check the Staking Portal for current terms.

Who is behind Cartesi

Erick de Moura, Augusto Teixeira, Colin Steil and Diego Nehab co-founded Cartesi in 2018. That comes from secondary profiles, not a primary Cartesi page. Cartesi's September 2026 update names Erick de Moura as a co-founder who led the website redesign. The project is stewarded by the Cartesi Foundation.

Risks and open questions

• Adoption. The Foundation itself says technical merit did not drive adoption as it expected. The October 2026 shift ties much of its effort to one independent commercial product it does not control.
• Maturity. The core v3 contracts and the fraud-proof system are still in alpha, with no new mainnet deployments in Q3 2026.
• Honeypot limits. The Stage 2 rating applies to the Honeypot deployment, which holds a small value and has a single withdrawal address. It is not a rating of production applications.
• Token economics. Any change to CTSI's utility would go through a separate governance process that has not been announced.

Limitations: the technical and roadmap details are Cartesi's own statements, and only the Honeypot's status was checked against an independent source (L2BEAT).