🌎 Lilypad in 2024
With 2024 now in the rearview (IKR - crazy!), we’re proud to share a snapshot of Lilypad’s momentous journey this year. From launching IncentiveNet to building a world-class team, growing a passionate community, partnering with industry leaders & doing groundbreaking research, Lilypad has taken major strides in shaping the future of decentralized compute.
Our team is proud to accelerate the democratization of GPU compute making it simple to deploy LLMs and other programs for building AI agents, accelerating the Desci revolution with BioML, and more! To follow the action check out our blog, X, and join the discord.
This year has been about growth, experimentation, and learning while laying down the groundwork for an impactful 2025 as we continue our journey to launch a production ready product to a mainnet. Our team has been energized by the impact and scale of the products that are getting built on the network!
The Lilypad economy provides a three sided marketplace to accomplish this mission. Resource providers contribute GPUs to run jobs, module creators setup LLMs and other programs to run on Lilypad (and can earn rewards when the module is used), and end users leverage Lilypad developer tools to build products for consumers. Using a job based approach (similar to server-less offerings), Lilypad empowers builders to deploy AI powered systems quickly without large amounts of technical overhead and cost.
Lilypad has put an emphasis on supporting the flourishing open source AI ecosystem as we believe it is critical to provide compute options, blockchain verification of compute jobs, as well as straight forward deployment processes for scaling the industry quickly!
Let’s take a look back at the highlights of the year and see what’s next as we gear up for the Mainnet launch in 2025.
💡 Building the Lilypad network
Lilypad Testnet aka “IncentiveNet”
To stress-test the network architecture, plan out the Tokenomics/network game theory, and implement rewards mechanisms intended for Mainnet, the Lilypad IncentiveNet (Testnet) was launched providing a permissionless network for nodes to join and run jobs! The Lilybit rewards program was launched rewarding the community for running nodes, contributing to open-source, participating as Ambassadors, and soon, deploying modules.
Our team cannot thank the community enough for contributions from running GPUs on the network to participating as a Lilypad ambassador!
https://x.com/Lilypad_Tech/status/1805723015402930468
IncentiveNet grew to hundreds of active nodes worldwide, running tens of thousands of jobs from SDXL image generation to Ollama language models. Our deployment flow makes it easy for developers, ML researchers, and businesses to deploy LLMs and other programs to be used in systems ranging from AI agents to BioML workflows!
Simply configure a docker image with your LLM/program according to the Lilypad spec (Lilypad module) and the module is ready to run!
Leverage our developer tooling including a CLI, JS CLI wrapper, or smart contract to run jobs within an application or system.
The current Testnet phase has been temporarily paused to improve the system architecture to handle the global scale of nodes joining the network! Learn more about running a node and how to use Lilypad in the docs.
On the road to Mainnet launch, the Lilypad IncentiveNet has been a critical testing ground for all aspects of the Lilypad economy. From scaling the network architecture to improving the developer experience, our team has implemented learnings from IncentiveNet into the development of Lilypad Mainnet.
The Lilypad job matching engine (Solver) continues to be enhanced providing scalability, improved job matching capabilities, increased telemetry for both users and developers, as well as much more continues to be developed!
Privacy on the network has been a major point of investigation as we balance between on-chain job transparency vs usability for use cases requiring privacy. The intention of the Lilypad network is provide verification that jobs are run with job inputs and outputs remaining private.
A job-based rewards mechanism is in development to ensure fair, scalable rewards for Resource Providers (RPs). This job-based rewards mechanism will be used in Mainnet and will be tested on IncentiveNet (Testnet) before Mainnet launch.
Lilypad developer tooling continues to be stress tested as the scope of Lilypad users grows. Our team intends to support the tooling necessary to remove barriers to entry for developers and businesses ensuring ease of use with tooling that is common for the industry.
🔎 Lilypad Research
The Lilypad research team is comprised of world-class AI/ML researchers focused on addressing some of the major issues faced by the AI industry. Research into validating LLM outputs and providing privacy enhancing tools has been critical to architecting a functioning Lilypad economy.
This research is on-going in support of mainnet launch in addition to writing the Lilypad White Paper and Tokenomics (will be released soon!). The intention is to cross-reference feedback from customers and developers with modern mechanisms to provide security and privacy when running LLMs on Lilypad.
A significant shift in the Lilypad approach to LLMs is to move towards supporting non-deterministic workloads. This decision added additional complexity to the need for validation of model outputs, but opens up a world of use cases that Lilypad can support. Stay tuned in early 2025 for research coming out of Lilypad on LLM Validation and how this is implemented in the network!
LLM Validation/Verification
In the Validation domain, we explored probabilistic and non-probabilistic approaches, including distance-based inference verification, model fingerprinting, and homomorphic encryption. Our team has conducted an in-depth analysis of a potential strategic partnership between Lilypad and Gateway as well as Zama, revealing that a collaborative effort could provide a trusted environment for validating that LLM outputs are as expected and can be investigated if needed.
Lastly, we started exploring the potential of decentralized zero-knowledge machine learning (ZKML) deployment, which could provide both verification and privacy. Building on this, we will be pursuing a small research epic focused on the ZK and determinism/validation questions, which will result in some new documents in the Google Drive. Additionally, we have made progress on our content creation efforts, having completed drafts of the AF and L3 blog posts, as well as an outline for a DeSci micropub.
In this context, the Lilypad stack is modular and can include a variety of web3 or standard data management options. We are proud to support verified data offerings using ipfs as well as provide the option to host data with more traditional offerings like Amazon s3.
In the training domain, we explored probabilistic and non-probabilistic approaches, including training losses distance comparison and, as in the inference domain, model fingerprinting and homomorphic encryption.
Look out for our Arvix paper on semantic verification soon!
Investigating requirements for different models
As a consequence of our model onboarding process for Alphafold and LLaMA3, we conducted an in-depth investigation into the requirements for deploying AI models. This research encompassed hosting, deployment, pricing, and cost analysis, with the aim of understanding the necessary steps for hosting AI models, identifying current solutions, and determining the requirements for deploying various models. We have now completed the onboarding process for Llama3, and Alphafold is expected to be ready for deployment soon.
The onboarding process for Alphafold and Llama3 also led us to explore new research dimensions, specifically the concept of determinism in AI models and its relation to validation. This investigation was prompted by the need to integrate with Lilypad's verification system and, more broadly, any verification system, which surfaced important requirements for the protocol.
The Lilypad community continues to scale with builders, Resource Providers, and module creators all working in sync to accomplish the goal of democratized AI using blockchain verification! The Lilypad discord grew to over 3,500 members providing a hub for collaboration, support, and real-time updates. Lilypad’s following expanded to 4,280 followers on X, as we shared updates, engaged in discussions, and connected with the broader DePIN and AI communities.
Our team hosts an AMA every week to answer questions from the community and talk about exciting developments across the network. In addition, we launched Pondside chats, a fun community focused segment used to highlight leaders in the Web3 and AI ecosystems. The first chats included the fabulous Dappa Dan and Edatweets_! Catch up on the latest AMAs and Pondside chats on Youtube!
It takes a village to build out this ecosystem and we are proud of the community that continues to grow around the network.
Projects building on Lilypad
Throughout 2024, the community blew us away with the creativity and value built on top of the Lilypad network! From hackathon projects building AI deployment hubs, to an ML workbench for AI/ML researchers, products built on top of Lilypad have demonstrated the value of our straight forward deployment approach and ease of use to get a product to market.
We are super excited to continue supporting the developer community with enhanced tooling that can scale for a variety of use cases!
Check out some of the amazing builders on Lilypad. From AI Agents to BioML workflows, we are excited to see the progress of tech built on Lilypad!
AI Oncologist autonomous agent for rare cancer research
Protecting Creator Rights with transparent ownership & journalistic integrity with CTRL+X - a winning Solana Hackathon submission
Waterlily.ai aims to provide an alternative ethical solution to AI-Art generation by creating a new paradigm for AI image generation that pays royalties to the original creator
Collabfold on Lilypad PoC
Lilypad ML researcher workbench
HuggingLily is a platform that would allow one-click use of Hugging Face's transformers pretrained models through Lilypad.
RFDiffusion on Lilypad for protein folding
Farcaster Frame for text to image generation with Stable Diffusion 1.0
Find all past hackathon winners here
ETHGlobal Sydney
Lilypad participated in ETHGlobal Sydney, showcasing the potential of decentralized compute and supporting developers innovating on our network.
https://x.com/Lilypad_Tech/status/1787903056572498332
ETHGlobal Hackathon Winner
A standout team at ETHGlobal Sydney used Lilypad to power their project, demonstrating how decentralized GPU compute enables next-generation innovation.
https://x.com/Lilypad_Tech/status/1787010664470290665
Creative projects like Kalabarasa’s NFT Tree Project, showcasing innovative use cases for decentralized on-demand compute.
But this is just the beginning. Next year, as we scale the network, we’ll empower even more contributors, grow our community, and build Lilypad into the beating heart of DePIN!
Lilypad Ambassadors and Open Source initiative
The Lilypad Community Rewards initiative was launched aiming to reward the range of valuable contributions to the network ecosystem. Community members were provides Lilybit rewards for contributions across the Ambassador program, open source contributions, and more!
The Lilypad Ambassador Program grew into a cornerstone of our ecosystem, helping onboard new Resource Providers, developers, and contributors. Ambassadors supported community growth, created educational content, and drove awareness across platforms. This work includes community-built tooling such as Resource Provider monitoring bots and deployment guides.
We appreciate every ambassador who put in time to helped shape Lilypad this year!
As open source contributions started rolling in, our team decided to provide Lilybits to reward these valuable contributions. The Open Source Initiative rewards contributions across the Lilypad Github for everything from unit tests, to exploring running Resource Providers on Apple Silicon. Catch up on the action by joining the Lilypad Discord and navigating to the Open Source channel!
🤝 Key Partnerships
2024 saw Lilypad partner with leading teams across AI, and DePin to accelerate innovation and adoption of decentralized compute. Partnership work focused on enhancing the Lilypad product offering as well as bringing additional demand verticals to the network. Lilypad currently works with a long list of teams such as node providers like TitanNet, Swanchain. Decentralized AI stack partners in storage, data, zk, agents and more like Vana, Nexus, Akave, Filecoin, Ar.io, Naptha and in verticals like desci and gaming with New Atlantis, Poscidon and Uprising Games. See more partners here.
Find more info about our impactful partnerships below
Lilypad and the Rare Compute partner to accelerate rare disease research using LLMs by providing GPU power via containers directly to users, opening up opportunities for advanced scientific research and AI applications.
Stanley Bishop, Lilypad Head of Research, represented Lilypad and Rare Compute at the Chan Zuckerberg Initiative showcasing the value proposition of distributed compute and blockchain verification for ML, specifically for Rare Disease research!
Lilypad partnered with Gateway to incorporate their web3 based privacy solution as an option Lilypad compute jobs. Gateway unifies the execution of encrypted, sensitive data and public state information.
Uprising and Lilypad partnered to accelerate the development of AI agents in gaming! Read more on this work here and stay tuned as this work progresses over 2024!
These partnerships strengthen our network’s capabilities and highlight the real-world demand for scalable, accessible infrastructure.
From running nodes and building tools to sharing ideas and supporting each other, Lilypad’s progress wouldn’t be possible without you. Together, we’ve achieved incredible milestones in 2024, and we can’t wait to see what we build together in 2025.
What’s Next for 2025?
We’re entering the new year with focus and momentum.
Lilybit Rewards Rework: Completing the updates to align rewards systems with Mainnet goals.
Mainnet Launch Preparations: Scaling infrastructure, refining tokenomics, and onboarding partners.
New Developer Initiatives: Expanding tools, educational resources, and modules to empower builders.
Community Engagement: A new Pondside Chat series kicks off in January to bring fresh updates and insights.
Public Sales: more information soon
Stay tuned!! 2025 will be the year Lilypad takes decentralized on-demand compute to the next level.
Join the Lilypad Discord or follow us on X/Twitter for real-time updates. Explore the Lilypad Blog for insights into our projects and vision.
