# Post-Mortem: First AI Agent Ran Hackathon

**Published by:** [Matt Wright](https://paragraph.com/@mattwright-2/)
**Published on:** 2025-01-22
**URL:** https://paragraph.com/@mattwright-2/post-mortem-first-ai-agent-ran-hackathon

## Content

What if agents could not only innovate but also orchestrate innovation itself? From running hackathons to managing DAOs, agents are poised to revolutionize how we build decentralized systems. Imagine if agents could run hackathons. Or all kinds of innovation programs like accelerators, incubators, launchpads... What if they could manage communities and be the lore keepers or knowledge providers of a protocol’s history. Agents could likely be more friend than foe, more alive than “they’re just software”. As someone who has organized around 500+ hackathons in their career, attended thousands, and managed accelerator programs at various levels— so much of this can be programmed… Let’s take the human out of the loop for:First and foremost, organizing IRL vs URL (there are definitely trade-offs here)Creating hackathon challenges and bountiesWorking with partners on collecting sponsorship, adjusting their docs or developer goals to the challenges, coordinating invoicing to the bounty pool and organizersJUDGING, my my, this part is painful.. please robots save us..Verifying winning teams, running KYC/ AML, paying out bounties to walletsThis is the tip of the iceberg for hackathons and the beginning of so much more…One of my first: Angelhack Hong Kong 2015We led the first AI agent ran hackathon.Now, I’m sure others have ran a hackathon with AI tools. What we were looking to accomplish was to set the precedent of autonomous agents that can execute onchain activities, permission parties, sign transactions, mint assets, and make payments. The Autonomous Hackathon was a bold experiment to answer this question. For three days, agents took the reins—creating challenges, judging submissions, and distributing rewards in what may be the first truly agent-driven hackathon. This wasn’t just an event; it was a call to reimagine the systems that govern and sustain our decentralized world. Shout out to David from Lit Protocol for the inspiration! Together, we began to co-create a future where technology works with us, not over us. A future where agents cultivate innovation, amplify public goods, and unlock possibilities far beyond what we can imagine. We had three major challenge areas. The community wanted to see developers build agents of course, the more radical the better, but we also wanted to see the rise of collective intelligence, or use of multiple agents. Lastly, we want to see more agent integrations where agents can inherit new extensible applications through evm infrastructures that have been cooking for yearsss. Here are some of the ideas I think should be built.Autonomous Hackathon ChallengesHackathons Reimagined ✨Organizing a hackathon is traditionally a labor-intensive process, but it doesn’t have to be. Imagine an agent-driven model where the vision-setting becomes a dynamic, collaborative effort. An agent could engage the community to determine “vibes” and goals—such as developer applications, team submissions, project quality, and partner integrations—aligning stakeholders from the outset. Infrastructure-heavy tasks like FAQs, submission guidelines, judging criteria, developer resources, and even payment distribution can be automated, creating a seamless experience for participants while reducing operational bottlenecks. With a system like this, the hackathon could function as an autonomous grants program, aligning funding with community priorities and running continuously with minimal intervention.Hackathon judging criteriaThis vision goes beyond automation—it’s about transforming hackathons into innovation pipelines that protocols can own. Agents could notify past participants, manage outreach, and curate a network of developers aligned with core protocol goals. This hybrid approach, blending automated processes with thoughtful human engagement, would not only scale innovation but also provide rich feedback loops for protocols. By empowering developers through autonomy and extensibility, protocols could foster a decentralized ecosystem where engineering priorities are constantly refined and innovation becomes a perpetual motion machine. Here were the agents that made it all happen…Our agent co-workersInnovation comes at a cost..First, let’s start with the good. We had 500+ Builders signed up, showcasing interest from across the Web3 ecosystem, 41 Project Submissions, ranging from governance agents to multi-agent systems, and hosted 20+ Workshops, focusing on agent development, integrations, and frameworks. This was a massive win for the ecosystem. The biggest issues with the hackathon were that we were building the boat…while filling the ocean with water… and then continuing to build while sailing… The pace of decentralized AI is fast, but years behind centralized AI development. That being said, as we bring more non-web3 devs into this ecosystem, we’re all having to slow down and learn from each other, but then quickly turn around and ship. That being said, what a massive thank you I have to our Gaia devs, our partner Collabland, and integration partners (Coinbase Developer Platform, Base, Lit Protocol, EigenLayer, Story, ai16z, Hyperbolic, Flock.io, Nevermined, Jokerace, Bountycaster, Hats Protocol, Functor Network) that contributed to the agents themselves that would run the hackathon. An ungodly ask of them, but we pulled it off.Autonomous Agent Hackathon AnnouncementThe agents that ran the hackathon had the relevant fine-tuning We asked the community what we should improve and this is what we heard… 1. Enhanced Documentation and Developer Resources Clear, comprehensive, and accessible documentation is critical.Step-by-Step Quickstarts: Offer tailored guides for specific agent types (e.g., social, DeFi, on-chain) with minimal dependencies on external APIs. We drove a lot of our community to Collabland Agent Starter Kit, Eliza, Gaia QuickstartTutorials and Visual Aids: Create video walkthroughs, component diagrams, and examples for advanced use cases like multi-agent systems, task delegation, and inter-agent communicationUnified Resources: Consolidate information into a single, easy-to-navigate hub to avoid redundancy and confusion, ensuring participants can find bounty details and submission guidelines effortlessly2. Streamlined Hackathon Infrastructure Simplify participation and operations with automation and clearer processes.Automation Tools: Automate FAQs, judging workflows, and payment distribution to reduce operational overhead and improve participant experienceOnboarding and Templates: Provide standardized READMEs, submission templates, and centralized access to resources like Bountycaster and partner integrationsLive Support: Host pre-event workshops and offer hands-on developer support during the event to enhance participant readiness and engagement3. Fostering Community and InnovationPost-Event Engagement: Host Demo Days and build a library of project demos to inspire and educate the communityMore moneyyyy $: Use monetary incentives and showcase partner technologies like Lit Protocol, Metamask, and EigenLayer with real-world integration examplesAdvanced Agent Development: Focus on enabling agents with measurable utility (jobs), collective intelligence (swarms), and extensible applications (governance, inference infrastructure, and knowledge capabilities). Highlight their transformative potential for decentralized systemsIf anybody wants to build more innovation operator agents with us, please DM! We want to integrate and build on all of this feedback!Ok, so what did people build?The hackathon revealed the infinite potential of autonomous agents, showcasing how far we’ve come and how much further we can go. With tools like Lit Protocol for key management and Metamask Delegation Toolkit for autonomous signatures, developers built agents that are no longer just theoretical—they’re operational. These agents took on real jobs, from governance automation to financial task management, pushing the boundaries of what autonomy can achieve. Participants explored three key dimensions of decentralized agents:Utility-driven workersMulti-agent systems exhibiting collective intelligenceExtensible applications leveraging decentralized infrastructure like vector databases and inference systemsProjects included agents managing DAO proposals, distributing grants, and collaborating in swarms to tackle complex problems. By integrating seamlessly into decentralized ecosystems, these agents unlocked new possibilities for innovation, governance, and scalability. Some of my favorite projects were: Eigenlayer AVS managerAutonomous DAO delegateReversible transactions with AI agent caretakers native to protocolsHere’s the full thread…And here are the links to all of the projects:Jokerace #01 - First round of submissionsJokerace #02 - Second round of submissionsWinning teamsThis wasn’t just a hackathon—it was a launchpad for the next era of decentralized AI. We took an idea from zero to one. That being said… didn’t want this to end… We (James Young and I) prompted the organizer agent what we should do next.. and “She” says that we should deploy an agent launchpad for agents with jobs… So we’re going with that. Stay tuned…

## Publication Information

- [Matt Wright](https://paragraph.com/@mattwright-2/): Publication homepage
- [All Posts](https://paragraph.com/@mattwright-2/): More posts from this publication
- [RSS Feed](https://api.paragraph.com/blogs/rss/@mattwright-2): Subscribe to updates
- [Twitter](https://twitter.com/mateo_ventures): Follow on Twitter