In February 2025, a momentous event unfolded at the crossroads of two of the most exciting trends in blockchain today: artificial intelligence and decentralized finance. Known as the EigenLayer AI-DeFi Hackathon—or Experiment Week #3—this invite-only gathering ran from February 10 through February 16, 2025. It was a joint initiative between EigenLayer, a leading restaking protocol on Ethereum, and Cartesi, the team behind a Linux-based virtual machine designed for blockchain development. From the moment the hackathon was announced, it attracted attention not only because of the reputations of its organizers but also because it promised to explore uncharted territory: building decentralized applications that combine the analytical power of AI with the financial innovation of DeFi.
At a time when many blockchain projects remain siloed—AI work often happens off-chain in centralized data centers, and DeFi protocols focus mainly on token swaps, lending, or synthetic assets—this hackathon set out to bring these two worlds together. Participants were asked to imagine applications where AI models don’t just run in isolation but can be made verifiable and trustless by blockchain. Picture an automated credit-scoring system whose data feeds and decision logic are recorded immutably on Ethereum, or a trading bot that uses machine learning to forecast price movements and whose every trade can be audited on-chain. These possibilities drive toward a future where “killer apps”—applications so compelling that they spur mass adoption—emerge naturally from the synergy of AI and DeFi.
To understand why this hackathon carried so much weight, it helps to step back and consider the histories of the two platforms at its core. EigenLayer arrived on the scene with a single bold idea: restaking. On Ethereum, validators lock up—stake—their ETH to secure the network. EigenLayer allows those same validators to “restake” that ETH in support of other networks or services, effectively reusing capital to secure multiple protocols. Meanwhile, Cartesi introduced a Linux-based virtual machine that runs complex computations off-chain in a familiar environment. Together, these two technologies promised to offer hackathon participants the best of both worlds: a way to build sophisticated AI models without being constrained by typical on-chain resource limits, plus a means to secure their applications using Ethereum’s vast staking ecosystem.
Over the course of seven intense days, participants collaborated with mentors from both teams, joined workshops on topics ranging from smart-contract security to AI model validation, and raced to deliver working prototypes. By the end of the event, a handful of standout projects had emerged—proof-of-concepts that hinted at a future in which AI-powered DeFi solutions are more transparent, trustworthy, and accessible than ever before. On the final Demo Day, each team had just one minute to show how their application worked and to explain why it deserved a share of the $12,000 prize pool.
This article pulls together every detail of the hackathon—from its ambitious objectives and carefully crafted structure to the underlying technologies and the ideas that participants explored. We’ll look at what made EigenLayer’s restaking so transformative, why Cartesi’s Linux VM matters, and how the concept of verifiable AI has the potential to change the way we think about automated decision-making. We’ll dive into the potential use cases that teams tackled, consider the broader implications for the blockchain ecosystem, and reflect on how this event might set new standards for innovation. By weaving together firsthand accounts, technical explanations, and thoughtful analysis, we hope to give you a granular, behind-the-scenes look at how EigenLayer’s AI-DeFi Hackathon laid the groundwork for a new generation of decentralized applications—applications that could very well become the “killer apps” that drive mainstream adoption in the years ahead.
Background on EigenLayer and Cartesi
EigenLayer: Extending Ethereum’s Security Through Restaking
At its core, Ethereum relies on a network of validators who stake ETH to secure the blockchain. In return for locking up their capital and running validator nodes, these participants receive ETH rewards. This staking model has proven remarkably effective at maintaining Ethereum’s security and decentralization. EigenLayer’s foundational idea was both simple and revolutionary: allow existing Ethereum validators to “restake” their already staked ETH toward new services and protocols without having to set up separate validator fleets for each one.
By enabling restaking, EigenLayer creates a network effect. Each additional protocol that opts into EigenLayer’s shared security can tap into the same pool of capital that Ethereum uses. That means new networks or services no longer need to recruit their own set of validators or require fresh capital for securing the chain. Instead, they “plug into” the security guarantees that Ethereum already enjoys. In practical terms, this considerably lowers the barrier to launching new projects—especially for experimental or early-stage networks that might otherwise struggle to attract a dedicated validator community. Developers can build with the confidence that their application benefits from Ethereum-level security even before it has any significant user base.
By early 2025, EigenLayer had locked up over $9.7 billion worth of ETH through restaking. That impressive figure reflected growing trust among validators in EigenLayer’s architecture and the perceived value of offering staking services beyond Ethereum itself. As more projects joined the EigenLayer ecosystem, the combined security pool grew, making it even more attractive for new entrants to tap into.
EigenLayer also built in incentive mechanisms and safeguards to maintain network integrity. In April 2025, the project deployed its “slashing” feature. In simple terms, slashing is a penalty mechanism: if a validator behaves maliciously—by double-signing, going offline during consensus, or otherwise failing to meet protocol requirements—their staked ETH can be partially or fully confiscated. Integrating slashing was crucial because, with restaking, you had Ethereum validators securing multiple services simultaneously. If one of those services behaved badly, there needed to be a way to hold validators accountable. The April 2025 slashing rollout completed EigenLayer’s original vision by ensuring that malicious or negligent behavior would be met with real financial consequences. Validators now had a strong disincentive to misbehave, keeping the shared security model robust.
Cartesi: Bridging Traditional Software Development and Blockchain
On the other side of the hackathon’s technological equation lies Cartesi. The biggest challenge for many developers entering the blockchain world is the need to write smart contracts in domain-specific languages—Solidity, Vyper, or Rust for various chains—and adhere to strict on-chain resource limits. Complex computations, large data sets, or even the use of libraries common in mainstream software development can be prohibitively expensive or simply infeasible on-chain.
Cartesi set out to eliminate those constraints by creating a Linux-based virtual machine (VM) that runs off-chain but remains verifiable on-chain. Its architecture is built around RISC-V, a lean, open-source instruction set tailored for flexibility and efficiency. Developers can write code in any language supported by Linux—C++, Python, even full-blown AI frameworks—and run those computations inside Cartesi’s VM. The VM handles all the heavy lifting off-chain, then produces cryptographic proofs that the computation was carried out correctly. Those proofs and final results are then committed to Ethereum (or other blockchains), where smart contracts can verify them. The outcome is that developers no longer need to rewrite complex algorithms in Solidity or worry about gas costs exploding because they’re crunching large datasets.
For a hackathon centered on AI and DeFi, Cartesi’s off-chain computation model was especially powerful. AI models often require matrix multiplications, large-scale data processing, and iterative training—all of which are impractical to run on Ethereum mainnet. Cartesi’s Linux VM let participants build AI routines locally in a conventional environment, test thoroughly, and only push proofs on-chain. This removed a massive barrier: it meant that an AI-driven DeFi product could maintain a high level of computational complexity without incurring prohibitive gas fees or compromising security.
The Synergy: A Unified Platform for Innovation
When you put EigenLayer and Cartesi together, you get a synergistic platform. EigenLayer ensures that the validator staking capital behind each project is plentiful and already battle-tested on Ethereum. Cartesi, meanwhile, guarantees that the computational side—particularly AI routines—can run off-chain in a fully flexible environment yet remain verifiable. For hackathon participants, this meant they could think bigger than ever before. They could prototype AI-driven financial tools that pull in large datasets, process them using traditional ML frameworks, and rely on the blockchain only for final proofs and on-chain state changes.
In other words, instead of worrying about how to secure their code or how to handle complex math on-chain, developers could focus on crafting the user experience, refining AI model accuracy, and designing novel DeFi mechanics. EigenLayer took care of security; Cartesi took care of computation. The result was an infrastructure designed to let creativity flow. Throughout Experiment Week #3, mentors from both EigenLayer and Cartesi emphasized that participants should treat security and computation as solved problems and push the envelope in combining AI with financial primitives.
The Synergy of AI and DeFi in the Hackathon
Verifiable AI: Establishing Trust in Automated Systems
One of the central themes of the hackathon was the concept of verifiable AI. In traditional systems, AI decisions—such as loan approvals or medical diagnoses—often lack transparency, leading to trust issues. By leveraging blockchain’s immutable ledger, developers aimed to create AI systems whose outputs are not only transparent but also verifiable by all stakeholders. This approach ensures that AI-driven decisions can be audited, fostering trust among users and regulators alike.
In his keynote, Nader Dabit, Director of Developer Advocacy at EigenLayer, emphasized: “AI is powerful—but verifiable AI is transformative. These are systems where the blockchain app isn’t the product; it’s the trust layer making AI reliable enough to redefine how society interacts with tech.” By integrating AI with blockchain, developers can create applications where decisions are not only automated but also transparent and accountable.
Potential Use Cases
The hackathon encouraged participants to explore various applications that combine AI and DeFi, including:
AI-Powered Trading Bots: Developing bots that utilize AI algorithms to analyze market trends and execute trades, with all actions recorded on the blockchain for transparency.
Decentralized Credit Scoring: Creating AI models that assess creditworthiness based on blockchain-verified data, enabling fairer lending practices in DeFi platforms.
Automated Insurance Claims: Implementing AI systems that process insurance claims efficiently, with decisions and data stored on-chain to ensure accountability.
Infrastructure Enabling Innovation
The synergy between EigenLayer and Cartesi provided the necessary infrastructure for these innovations. EigenLayer’s restaking protocol enhances security and scalability, allowing developers to build robust applications without establishing their own validator networks. Cartesi’s Linux-based virtual machine enables the execution of complex computations off-chain, facilitating the development of sophisticated AI models within dApps.
This combination allows developers to create applications that are not only powerful and efficient but also trustworthy and transparent, addressing some of the most pressing challenges in the current technological landscape.
Notable Projects and Outcomes
Standout Projects and Their Core Innovations
While the hackathon organizers did not publicly release a full list of winning projects and their code, a few themes and examples emerged from team presentations and post-event write-ups. Collectively, these projects illustrate the practical potential of blending AI with DeFi underpinned by secure restaking.
AI-Driven Risk Prediction for On-Chain Insurance: One finalist team built a delegateable, modular system that leverages an AI model to predict the likelihood of a user submitting a valid insurance claim. The model was trained on a combination of on-chain data—such as transaction histories, oracle feeds for weather events—and off-chain data like satellite imagery or IoT sensor readings. Because much of this data is numerical and high-volume, the team used Cartesi’s Linux VM to run the neural network off-chain. Once the model generated a risk score, it posted a succinct proof to Ethereum. The associated smart contract then determined policy premiums dynamically. The more data a user provided—proofs of digital receipts, public data feeds—the more accurate the risk score became, ensuring fair premiums and reducing exposure for insurers.
Decentralized Autonomous Trading Fund (DATF): Another team created what they dubbed a “DATF,” or Decentralized Autonomous Trading Fund, effectively a mutual fund powered by AI. The idea was that investors could pool ETH into a smart contract managed by an AI model that continuously adjusts the fund’s allocation across multiple DeFi protocols—lending platforms, liquidity pools, yield farms—based on real-time market signals. The AI component consumed various data feeds: on-chain metrics (TVL changes, order-book depth), off-chain sentiment (social media, news), and macroeconomic indicators. Cartesi’s VM ran the model off-chain and generated proofs of each rebalancing decision. By posting those proofs, the smart contract could automatically execute swaps or reallocations on behalf of investors, while maintaining transparency about why each move was made. The result was a zero-trust investment fund that promised higher yields than passive staking but with a level of risk-management transparency unseen in typical DeFi pools.
On-Chain Credit Marketplace: A third project focused on enabling peer-to-peer loans in regions where traditional credit scoring is inaccessible. Users submitted on-chain attestations—proofs of stable income streams from decentralized gig-economy work, historical transaction data, or even cross-chain NFT ownership as a proxy for net worth. An AI model running in Cartesi’s Linux VM evaluated these attestations and generated an on-chain credit score. Lenders could then filter borrowers by score range and offer loans at interest rates dynamically determined by the model. Smart contracts managed collateral—often stablecoins or tokenized assets—locking them until repayment was complete. The project showcased how combining verifiable AI with on-chain finance could democratize lending in underbanked regions.
Impact Assessment
Though no single project solved every challenge, collectively these hackathon entries demonstrated that:
Verifiable AI Works in Practice: By using Cartesi’s off-chain computation model and posting proofs on Ethereum, each team proved that complex AI routines could be integrated into a DeFi pipeline without sacrificing transparency or security.
Restaked Security Lowers Barriers: None of the teams had to assemble a network of validators to secure their application. Instead, they tapped into EigenLayer’s existing validator set, receiving robust security guarantees from day one. That meant early-stage projects could be confident their code wasn’t vulnerable to simple 51 percent attacks or reorg manipulations.
DeFi Innovation Steps Up: Many DeFi applications today rely on oracles or centralized price feeds. The hackathon entries showed that AI-driven oracles—models that aggregate and cleanse data—can offer more nuanced, timely insights. Automated credit scoring or dynamic risk assessment are not just buzzwords; they’re functional prototypes that could reshape lending, insurance, and asset management.
Educational Value for Participants: Even teams that didn’t win a prize walked away with a richer understanding of how to combine Web2 development practices with blockchain. Many participants had never heard of EigenLayer before and now had firsthand experience in using restaked capital. Likewise, Cartesi made it easy for traditional developers to embrace blockchain logic without wrestling with Solidity’s limitations.
Broader Implications for the Blockchain Ecosystem
Advancing Decentralized Applications
The hackathon underscored the importance of verifiable AI in building trust within decentralized systems. By leveraging blockchain’s immutable ledger, developers can create AI applications whose outputs are transparent and auditable, addressing concerns about the reliability and accountability of AI-driven decisions.
This approach has far-reaching implications for sectors like finance and healthcare, where trust in automated systems is paramount. For instance, AI-powered trading bots can execute transactions with a level of transparency that traditional systems lack, while decentralized credit scoring systems can provide fairer assessments by utilizing blockchain-verified data.
Encouraging Collaboration
The collaboration between EigenLayer and Cartesi exemplifies the benefits of cross-protocol integration in fostering innovation. By combining their respective strengths—EigenLayer’s restaking capabilities and Cartesi’s support for complex computations—developers were empowered to build sophisticated dApps that address real-world challenges.
This synergy highlights the importance of collaborative efforts in the blockchain ecosystem, where the integration of diverse technologies can lead to the development of applications that are more robust, scalable, and user-friendly.
Setting New Standards for dApp Development
The hackathon set a precedent for future events by demonstrating the feasibility and advantages of integrating AI with DeFi on a decentralized platform. It showcased how blockchain can serve as a trust layer for AI applications, ensuring that automated decisions are not only efficient but also transparent and accountable.
This paradigm shift has the potential to redefine user engagement in decentralized ecosystems, paving the way for broader adoption of blockchain technology across various industries.
The Dawn of Restaked ‘Killer Apps’?
The EigenLayer AI-DeFi Hackathon, held from February 10 to February 16, 2025, marked a significant milestone in the evolution of decentralized applications (dApps). By integrating EigenLayer’s restaking protocol with Cartesi’s Linux-based virtual machine, the event showcased the potential of combining artificial intelligence (AI) and decentralized finance (DeFi) to create transformative applications.
One of the key takeaways from the hackathon is the concept of verifiable AI. By leveraging blockchain’s immutable ledger, developers can create AI systems whose outputs are transparent and auditable, addressing concerns about the reliability and accountability of AI-driven decisions. This approach has far-reaching implications for sectors like finance and healthcare, where trust in automated systems is paramount.
The collaboration between EigenLayer and Cartesi exemplifies the benefits of cross-protocol integration in fostering innovation. By combining their respective strengths—EigenLayer’s restaking capabilities and Cartesi’s support for complex computations—developers were empowered to build sophisticated dApps that address real-world challenges.
Looking ahead, the success of the EigenLayer AI-DeFi Hackathon sets a precedent for future initiatives aimed at integrating AI and blockchain technologies. As the blockchain ecosystem continues to evolve, such collaborative efforts will be instrumental in driving the development of applications that are not only powerful and efficient but also trustworthy and transparent.
In short, the EigenLayer AI-DeFi Hackathon was not just a weeklong sprint for prize money. It was a clarion call announcing that the era of restaked AI-powered DeFi is upon us. From the first lines of code written in Cartesi’s VM to the final slashing safeguards enforced by EigenLayer, this event painted a vivid picture of what blockchain can become when trust, computation, and capital security are all woven together. As we continue down this path, the “killer apps” that emerge—those that draw in mainstream users through compelling utility—may well be the offspring of the ideas born during that pivotal week in February 2025. The dawn of restaked killer apps has arrived, and Experiment Week #3 showed us just how bright that future can be.
