AI is in full swing, but its foundation—data—is largely centralized, creating bottlenecks & raising critical questions about ownership & provenance. The Filecoin Hackathon, hosted by Encode, challenged builders to envision & construct a more equitable AI ecosystem powered by decentralized infrastructure.
Over 550+ builders hacked with Filecoin for 4 weeks straight birthing 86 AI powered projects, competing for a $77k Prize Pool, and ideas that could reshape the future of decentralized intelligence. They grappled with how Filecoin's immutable storage and growing capabilities, like hosting models such as Deepseek, can democratize AI's future.
The core mission was clear: leverage decentralized technologies to tackle AI's most pressing data challenges. Builders were tasked with addressing:
Data Provenance: Creating systems for transparently tracking data lineage and verifying untampered LLM models in an era rife with AI-generated content.
Ethical Data Sourcing: Designing solutions for data marketplaces and DAOs that respect user privacy while incentivizing authentic contributions, especially as high-quality data becomes scarcer.
Fair Attribution: Building mechanisms for transparent value tracking and equitable compensation for data creators, potentially through Data DAOs.
Efficient AI: Exploring how Filecoin's decentralized storage can enable more efficient and environmentally conscious model training and deployment.
A strong emphasis was placed on projects considering modular architecture, interchain interoperability, and the burgeoning agentic economy. As with a few articles from hackathons i’ve had the honour of participating in before, here is a selection of standout projects, particularly those recognized in Filecoin's main bounty track, "Hacking the Data Layer for AI," and other impressive finalists.
A small random selection of winners.
N&P - Agent Nexus Protocol
Product: https://github.com/Shubham-Rasal/Agent-Nexus-Protocol
Developer: Shubham Rasal
Promise: A unique way for AI agents to collaborate, addressing the difficulty of implementing multi-agent interactions and the lack of user-friendly UX for such systems. N&P features a multi-agent chat interface where agents, assigned by an intelligent task router (powered by Lilypad), can hop in to assist. Relationships between agents are defined by "agent networks" (e.g., for sales and marketing), influencing task distribution. Decentralized storage is used for agent memory and chain-of-thought.
Filecoin Track: 1st Place
My Take: N&P's approach to multi-agent collaboration through a familiar chat interface is smart. The concept of "agent networks" defining relationships & task routing seems powerful for complex workflows. Storing agent memory & thought processes decentrally is crucial for transparency & resilience. The live demo showcasing email outreach initiated via chat was compelling. This project hits the "agentic economy" theme & the judges rightly recognized its innovation & technical depth.
Datahive
Developer: Ademola Adebowale, Emmanuel Akanji
Promise: A token-incentivized data marketplace for AI. AI researchers can set up data collection campaigns (defining specs, duration, token rewards), and everyday users (data providers) can submit their real-world data (transaction records, sensor logs) to earn. Submitted data is verified using a Lilypad multi-agent workflow, with workflow details saved on Arweave. Datahive transforms raw contributions into AI-ready datasets, with Arweave as a retrieval layer for model training/inference. Smart contracts and Filecoin are used for dataset storage. Users can buy AI-ready datasets or sell their own for token rewards, all backed by Akave’s decentralized storage.
Filecoin Track: 2nd Place (Shared)
My Take: Datahive addresses the critical "ethical data sourcing" & "fair attribution" challenges head-on. The two-sided marketplace model is well-conceived, plus the use of multi-agent verification for data quality is a strong feature. Integrating Filecoin for storage as well as Arweave for retrieval/metadata showcases a robust, decentralized stack. This feels like a practical solution to a very real problem.
Decentralized Sign Language Translation For All
Product: https://github.com/ce20480/sl-teaching-website/tree/main
Developer: Aaron Vinod
Promise: Decentralized Sign Language Translation For All utilized a vision model to validate the quality of training images, measuring blurriness and content relevance, to create a dynamic ASL interpretation tool. The validated data is routed to a Data DAO on Filecoin, where contributors are rewarded for high-quality submissions, and stored on Akave for training and reward allocation.
Filecoin Track: Akave 2nd Place
My Take: This project stood out for its thoughtful use of decentralized storage, AI-driven validation, on-chain rewards, and real-world impact in accessibility and inclusion.
Frankie Agents
Product: https://github.com/Marshal-AM/franky
Developers: Gabriel Antony Xaviour, Sam Felix, Marshal Mathews
Promise: Monetize old devices (phones) by turning them into efficient AI agent hosts, tackling e-waste and the high cost of decentralized hosting. Old phones can run a quantized 3B parameter model locally. Device owners register their devices on-chain (Filecoin EVM). A checker subnet pings devices, handles consensus, and tracks uptime. Agent creators can deploy agents, choosing between on-device LLMs or Lilypad, with secrets encrypted via Lit Protocol. Each hosted agent's DNS is an ENS subdomain.
Filecoin Track: 3rd Place (Shared) & Akave 1st Place
My Take: The One Piece reference (Frankie rebuilding himself) was a masterstroke! Beyond the great branding, Frankie Agents presents a clever solution to both e-waste and the cost of AI compute. Enabling on-device LLMs on old hardware is ambitious. The on-chain registration and checker subnet provide a good framework for a decentralized physical infrastructure (DePIN) network. This project has a strong sustainability angle and a clear value proposition for both device owners and agent developers.
Other Notable Finalists & Ecosystem Winners:
Safe Idea (Finalist): Focused on creators securing, sharing, and selling digital products, using Lit Protocol for encryption, Storacha/IPFS for storage, and Filecoin for minting ERC-7155 tokens for access. Also featured an AI sales agent trained on low-res versions of ideas using Lilypad.
Deal Retrieve Checker (Finalist & Checker Network Winner): A way to get data retrieval guarantees on Filecoin. Users can raise disputes if retrieval fails, triggering a checker network to verify, potentially slashing a storage provider's stake.
Storage as Service (Finalist & Coop Hive Winner): A bridge between users and storage networks (Filecoin) verifying storage commitments and releasing payment only on fulfillment, using Filecoin storage obligations and attestations.
Dapper Diary (Finalist & Lilypad 3rd Place & Storacha Winner): A personalized AI research assistant (a decentralized NotebookLM). Upload documents, embeddings stored on Storacha, answers generated via Llama 3.1 on Lilypad, with chat history also on Storacha. Includes a podcast generation feature - built by dappadandev.
Kintask (Finalist & Recall Grand Prize Winner): Infrastructure to verify AI and computational work. A decentralized marketplace to request, execute, and verify tasks, using ZK proofs for answer validation and Recall for storing questions, answers, and evaluations.
Ktun (Finalist): An AI and blockchain-powered police complaint management system. AI analyzes voice complaints, routes them to officers, and stores all evidence (voice, photos, location) on IPFS/Filecoin.
Pulse (Winner): A solution for music creators to prevent pre-release leaks and manage AI/IP complexities. It uses random timelock encryption (potentially via @RandamuInc ) to secure songs until launch, NFTs and Filecoin for storage, Story Protocol for IP rights management, and smart contracts on Filecoin EVM (FVM). Music leaks are a persistent problem for artists. Pulse’s use of timelock encryption combined with Filecoin for immutable storage and Story Protocol for IP is a comprehensive approach. This directly addresses "data provenance" and "fair attribution" for creative works. The demo showing the countdown to release and integration with Story Protocol was clear. This has strong potential for real-world adoption by creators.
Cognivern | My Submission
For this hackathon, I built Cognivern, a decentralized platform for auditing, monitoring, and governing AI agents using Recall's blockchain-based storage infrastructure.
Twitter: @papajimjams
Promise: Cognivern aims to provide verifiable decision logs, an intelligence marketplace for reasoning patterns, governance protocols, and eventually, multi-agent collaboration. The MVP focused on agent activity monitoring (tracking decisions with persistent chain-of-thought logs on Recall), a basic reasoning pattern marketplace structure, & governance controls with policy guardrails. Building on Recall's decentralised storage.
Key Features (MVP):
Agent Activity Monitoring: Intercepting and logging agent reasoning steps to Recall, structured for efficient storage and retrieval.
Reasoning Pattern Marketplace: Foundational elements for listing and discovering reasoning patterns, with policy-based access.
Governance Controls: Policy enforcement (ALLOW, DENY, REQUIRE, RATE_LIMIT) with audit logging to Recall.
Recall Integration: Managing bucket creation, secure storage, and batch synchronization.
What worked well:
The integration with Recall for storing structured agent logs, policies, and metrics proved robust.
Designing the PolicyEnforcementService and AuditLogService to interact with Recall felt intuitive for creating verifiable trails.
The concept of storing chain-of-thought data in a standardized JSON format on Recall seems like a solid base for future verification.
What I'm working on:
Building out the front-end dashboard to visualize agent activity and manage policies more interactively.
Implementing the tokenized access control and subscription models for the reasoning pattern marketplace.
Exploring cryptographic proofs (like ZKPs) for verifying intelligence lineage and private attestations of agent behavior.
Refining the multi-agent collaboration protocols.
This hackathon underscored the critical need for robust, decentralized infrastructure as AI systems become more complex and autonomous. The focus on data provenance, ethical sourcing, and verifiability was evident in many of the top projects. Filecoin and its ecosystem partners like Recall, Lilypad, and Storacha are providing the essential building blocks for this next generation of AI applications. The emphasis on agentic systems and modularity points towards a future where AI is not only powerful but also transparent and accountable.
The quality of submissions was incredibly high, and it was inspiring to see so many builders tackling these difficult but crucial challenges. Huge lesson in how to develop a winning powerpoint presentation especially to tell the story of your product. Honoured to participate & win a prize!
Infrastructure Powering the Hackathon
The event was heavily supported by Filecoin and its broader ecosystem, including:
Filecoin / IPFS: For decentralized storage and data permanence.
FVM (Filecoin Virtual Machine): Enabling smart contracts and on-chain logic.
Partner Protocols: Checker Network, AAV (Arweave), Coup Hive, Recall, Random Network, Stroracha, LilyPad, providing specialized tools and bounties.
Encode Club: For organizing and facilitating the hackathon.
Workshops & Support
Throughout the hackathon, participants benefited from various workshops and support sessions hosted by Filecoin and partner teams, helping them get up to speed with the latest tools and best practices. The Grand Finale event was hosted by Rebecca (Encode), with judges including Sarah and Rishik (Filecoin/Protocol Labs), and partner representatives like Patrick (Checker), Bart (AAV), Levy (Coup Hive), Mark (Recall), the Randamu team, Jamie (Storacha), and Phil (Lilypad). Nikki (Protocol Labs) also shared insights into future builder programs like Modular Worlds and Founders Porge.
This Filecoin AIP Hackathon was a vibrant showcase of the innovation happening at the intersection of AI and decentralized technology. The projects demonstrated a clear understanding of the challenges and a creative application of the tools available to build a more open, fair, and verifiable AI future.
Supporting writers is beyond the call of duty, reading this far is its own form of glory, appreciate you.
Connect
Farcaster @papa — warpcast.com/@papa
Lens @papajams — lenster.xyz/u/papajams
Twitter @papajimjams — twitter.com/papajimjams
Read
PAPA: https://app.t2.world/article/clw3f0q9m24698020mcs051nt84
Akave: https://www.akave.ai/blog/akave-x-encode-how-builders-hacked-the-ai-data-stack
Over 100 subscribers
Discover the innovation from the Filecoin Hackathon where over 550 builders revolutionized AI with decentralized solutions. From ethical data sourcing to fair collaboration, projects like Datahive and N&P clinched top spots. Read more insights from @papa on this transformative event.