Cartesi x EigenLayer Experiment Week #3 Recap

The third edition of Experiment Week has come to a close! This time, we teamed up with EigenLayer to push and test the limits of the Cartesi Coprocessor alongside EigenLayer’s restaking protocol—unlocking new possibilities together.

Thank you to everyone who participated in this hackathon. From all the submissions, 13 finalists emerged, but only three could take the top spot.

Joint First Place

🥇 ThinkChain
An onchain service that enables smart contracts to perform verifiable LLM inference. Watch the DEMO here.

With support for popular open-source models like DeepSeek, Qwen, and SmolLM, ThinkChain can bring verifiability and trust to AI-driven on-chain games, AI-generated NFT metadata, AI-assisted token credit scoring, AI-generated DAO proposal summaries, and more.

Built by Guilherme Dantas, Eduardo Bartel, Felipe Grael.
GitHub repo here.

🥇 Cartesi Lido Oracle
A coprocessor that enhances the Lido protocol, one of the largest TVL protocols in the space, by replacing trusted parties with provable computation. Watch the DEMO here.

The project leverages Cartesi Coprocessor to compute reports from trusted relayers, which fetch data from Ethereum’s consensus layer, ensuring accuracy and resistance to manipulation.

By analyzing validator performance, staking rewards, and key metrics, it enhances Lido’s security and efficiency while providing transparent data for the broader DeFi ecosystem.

Built by Willem Olding.
GitHub repo here.

Second Place

🥈 PKMN.fun
Web3 Pokémon AI Battle Arena. Watch the DEMO here.

PKMN.fun brings competitive Pokémon battles into the web3 space by utilizing the Cartesi Coprocessor for decentralized battle simulations, ensuring that results are securely committed onchain.

Based on the widely used Pokémon Showdown simulator, the game opens up various opportunities, including betting and NFT integration. Season 1 features a demo on the Holesky Ethereum testnet, where teams are public, allowing challengers to strategically build unique teams to compete against the first player who joins. Play it here.

Built by Danilo Tuler.
GitHub repo here.

Crowd Favorite

Scribbl
Onchain AI tool that analyzes and ranks your sketching skills in real time. Watch the DEMO here.

The project addresses bias, centralization, and opaque decision-making by using a fully verifiable Linux machine for inference, ensuring unbiased and decentralized judgments.

With this on-chain coprocessor AI, the aim was to create a level playing field for doodlers—ensuring fair contests, transparent competition, and evaluations based solely on merit. This system aims to transform traditional, secretive AI evaluations into an open, verifiable, and enjoyable experience. Whether a doodle is a masterpiece or a merry mess, the decentralized judge is ready to review and reward every quirky creation! Test it out here.

Built by Shaheen Ahmed, Bruno Menezes, Helen Imah.
GitHub repo here.

NFTs

Onchain Fans
A decentralized platform that enhances fan engagement through the Cartesi Coprocessor, enabling transparent interactions between creators and fans while eliminating intermediaries in digital content sales and distribution. Watch the DEMO here.

OnChain Fans addresses pre-sale trust and transparency issues, where buyers are hesitant without assurance of receiving the full-quality version. The platform unlocks exclusive content, NFTs, and rewards via on-chain transactions, allowing artists to sell digital art and exclusive content while ensuring buyers receive the full-quality version once a pre-set goal is met.

Built by Joao Garcia, Zehui Zheng, Stephen Chen.
GitHub repo here.

Wartime Penguins
A Generative Art NFT collection on Holesky where all art is generated and published by a Cartesi Machine using the Coprocessor. Watch the DEMO here.

The project tackles the challenge of high-quality generative art NFTs, which often rely on centralized computation. By using Rust libraries, it generates a 3MB animated GIF NFT. Also, to prevent rug pulls, users can access all necessary IPFS blocks for independent art recreation. The Cartesi Coprocessor and Cartesi Machine ensure reproducibility, enhancing trust in the NFT ecosystem.

Built by Carsten Munk, Nyakio.
GitHub repo here.

Games

Escape from Tikal
A project that explores the use of trustless AI agents in cooperative games, where Player 2 is an LLM-powered AI, running entirely within the Cartesi Coprocessor. Watch the DEMO here.

By executing the AI logic off-chain but verifiably within the Coprocessor, this project ensures a transparent and fair competitive environment.

In tournament settings, all players interact with the same AI under identical conditions, preventing collusion and ensuring that no team gains an unfair advantage. The AI decision-making process remains verifiable and reproducible, strengthening trust in the game’s mechanics. During development, the team successfully ported PythonCartesi HLF to the Coprocessor, extending its capabilities and paving the way for broader applications beyond gaming.

Built by Carlo Fragni, Bruno Maia, Lyno Ferraz.
GitHub repo here.

Team Arthur/Trust Match
A decentralized matchmaking system for e-sports, built using the Cartesi Coprocessor. Watch de DEMO here.

The project ensures a transparent and verifiable matchmaking process by executing it within the Coprocessor, while traditional matchmaking systems, especially in ranked play, often rely on centralized algorithms that can introduce bias.

The system implements OptMatch, a generalized, data-driven matchmaking framework that requires minimal product knowledge, enabling fair, scalable, and decentralized matchmaking that can be integrated into various gaming products. The potential for this framework to be used as a Software as a Service (SaaS) further expands its applicability across different platforms.

Built by Arthur Vianna, Eduardo Loivos, Rayan Lima, ZacPrater.
GitHub repo here.

Cartesi Super Soccer
A project that empowers players to build NFT teams, submit lineups onchain, and earn XP after every match. Watch the DEMO here.

The Coprocessor swiftly calculates results and updates NFT levels in near real-time, eliminating slow settlement times with rapid base layer updates.

For instance, NFT XP stats adjust instantly after each match, ensuring immediate feedback and seamless progression. Beyond speed, the system unlocks advanced computation capabilities, enabling complex game logic and off-chain processing while preserving verifiable execution. Additionally, it offers a Linux-based environment, allowing developers to deploy and run computations efficiently using familiar tooling and infrastructure.

Built by Fabio Oshiro, Bruno Ochotorena.
GitHub repo here.

DeFi

CarteMarket
A multi-outcome prediction market leveraging the LSMR algorithm to calculate share prices more efficiently while minimizing computation costs. Watch the DEMO here.

CarteMarket offloads computationally intensive and costly operations, often requiring iterative cycles on traditional EVM-based systems, to the Cartesi Coprocessor. This transition makes prediction markets significantly more scalable and cost-effective. By handling complex calculations off-chain while ensuring on-chain verification, it enhances accuracy and accessibility.

To further optimize execution, the team is experimenting with aggregating user intents, allowing bulk execution of multiple market operations in a single transaction. This approach reduces transaction redundancy and improves efficiency in executing trades.

Built by Fabrizio Gianni.
GitHub repo here.

Intendo-64
A unique DeFi interaction platform that combines the nostalgia of retro gaming with modern DeFi capabilities and brings natural language processing to web3, powered by the Cartesi Coprocessor and EigenLayer AVS. Watch the DEMO here.

Intendo-64 allows users to execute complex DeFi operations using text or voice commands, enhancing the user experience while ensuring on-chain verifiability. It supports Uniswap for token swaps with slippage control, Aave for lending and borrowing with variable and stable rates, and general ETH/ERC-20 transfers.

The Cartesi Coprocessor processes user intents by analyzing natural language inputs, extracting DeFi parameters, and converting them into executable smart contract operations. The platform is built with React and TypeScript, featuring real-time contract event monitoring and a responsive retro UI.

Built by Suryansh.
GitHub repo here.

Social

Openquest
A platform that enables projects to engage their communities, foster growth, and reward meaningful contributions through verifiable quests. Watch the DEMO here.

OpenQuest provides personalized smart contracts for protocols, enabling no-code deployment of customizable, verifiable quests. Powered by Cartesi’s Coprocessor, it ensures trustless verification and fair scoring, regardless of participant numbers or quiz complexity. Rewards are automatically distributed based on the project model, allowing users to earn tokens for meaningful participation and improve their leaderboard standings.

The system also tracks developer contributions and community engagement through protocol-specific leaderboards, addressing challenges in tracking contributors and their engagement. This helps simplify the hosting of verifiable events like hackathons or grants, promoting genuine participation and sustainable growth in projects.

Built by Idogwu Chinonso, Hellen Stans, Evans Okoli, Samuel Babalola.
GitHub repo here.

Dev Tooling

MathCoprocessor
A simple solution that allows smart contracts to offload complex mathematical operations that are either too resource-intensive or unsupported. Watch the DEMO here.

This project leverages Cartesi Coprocessor to enable smart contracts to send an arbitrary Python mathematical expression along with a callback to be called with the ABI-encoded result. The implementation includes support for NumPy and arbitrary ABI encodings.

Built by Milton Jonathan.
GitHub repo here.

A huge thank you

To all the builders who participated. We’re grateful to have such a passionate community of builders who are willing to push themselves to keep creating something new. This Experiment Week surely won’t be the last, so stay tuned on our socials to keep up to date!

And a special thank you to the EigenLayer team for joining us on this journey—every integration brings us closer to making onchain adoption a reality for the masses!