🌎 Overview
Welcome to the latest Lilypad Project Report! We're excited to bring you updates on the strides we’re making to democratize access to GPU compute and build a robust, decentralized ecosystem that empowers users across the globe.
As the open source AI ecosystem flourishes, access to compute and developer tooling continues to grow in importance. ML researchers, developers, and businesses need low cost systems that can be relied on for creating, managing, and distributing models. Some open source models can run in relatively public settings without a need for privacy around outputs and associated data. Other workflows need privacy for the model data, inputs, and outputs. With a strong set of tools to implement ML models, applications can be built, research can be accelerated, and business outcomes improved.
To support this booming economy, the Lilypad Network provides a three sided marketplace for compute providers, model creators, and model users (researchers, developers, businesses, etc.) to run inference jobs. With compute jobs verified using the Ethereum blockchain (currently deployed to Arbitrum Sepolia L2) and model data/inputs/outputs saved with IPFS, the network provides a foundation for verification in AI processes and attribution for model creators. The Lilypad SaaS platform will provide an interface to this system for ease of use adding new models to run on Lilypad and developing systems using the models.
Currently in testnet, the Lilypad IncentiveNet aggregates idle GPUs from around the globe and runs containerized jobs on GPUs sourced from this network. Lilypad currently provides a CLI and smart contract tool set for developers to run jobs. Learn more by checking out the Lilypad docs and blog!
⚒️ Engineering Update
With IncentiveNet GPU capacity growing over time, updates continue to be pushed to testnet systems and infrastructure. Updates are generally pushed with a few of the core testnet goals in mind.
- Test Network Orchestration Architecture
By running thousands of compute jobs, IncentiveNet provides valuable insights into the scalability and efficiency of our system's orchestration, ensuring seamless operations as we move toward Mainnet. This work includes optimization of the Lilypad RP software stack, pre-flight RP settings, improvements to the market-matching engine (Solver), and more.
- Validate Game-Theoretic Mechanisms
We’re leveraging real-world participation to proof-test the economic and incentive structures that underpin the network. At the base level, we intend on incentivizing RP uptime and the ability to run Lilypad modules (AI workloads) including Proof of Works (Keccak) throughout the day.
- Deliver a Production-Grade Solution
IncentiveNet is not just a testbed—it’s a stepping stone to a fully operational marketplace for GPU compute, serving both supply and demand with reliability and efficiency.
Telemetry from RPs and the solver has been instrumental in guiding the architecture of Mainnet. By analyzing GPU performance, identifying root causes of job failures, and optimizing job-matching algorithms, our team uses this date to ensure Mainnet will scale seamlessly as the network grows.
New Release: v2.9.1 Fei-Fei Li (RPs update to the latest version!)
"The paradigm shift of the ImageNet thinking is that while a lot of people are paying attention to models, let's pay attention to data. Data will redefine how we think about models." ~ Fei-Fei Li
Lilypad v2.9.1 Fei-Fei Li is avaliable requiring all RPs to update! This update introduces a relatively simple version update with minimal changes affecting Lilypad Resource Providers. The primary update was in regards to improving how we log and report Solver connection errors.
🚨 Reminder: Every two weeks an older version of Lilypad is deprecated. RPs running the latest Lilypad version is crucial for the performance of the network. The two week window gives RPs time to update to the latest version. In this case, v2.9.1 was released, so v2.9.0 will be deprecated in 2 weeks. V2.8.0 was deprecated yesterday.
Announcements for RPs including v2.9.1 can be found in the updates-rp channel on Discord.
Lilybit_ Rewards Calculation Update
To enhance the fairness and efficiency of the Lilybit_ rewards mechanism within IncentiveNet, the Lilypad team has introduced several key updates to how GPUs earn rewards on the network. These improvements aim to ensure that contributions are recognized proportionally while maintaining a balanced incentive structure.
🔑 Key Changes:
GPU-Specific Base Points
Base Lilybit_ points are now determined by three key GPU parameters:
VRAM
CUDA Core Count
Tensor Core Count
This approach rewards GPUs optimized for AI workloads more accurately, ensuring fair compensation for their capabilities. The base points are multiplied by the daily rewards multiplier, calculated based on Proof of Work (PoW) uptime throughout the day. This encourages consistent participation while recognizing sustained efforts.
🔗 Learn More About the Multiplier
One GPU to one Resource Provider
Our system now requires 1 GPU for each RP to run PoW and jobs as well as earn rewards. This configuration provides the RP un-interrupted access to the server resources for running jobs. Industry standard uptime and job success rates are critical to servering Lilypad end-users with a high quality service.
Rewards Cap Adjustment
**To prevent an over-concentration of rewards among high-end GPUs and to support a wider range of contributors, a maximum of 20,000 Lilybits per day has been set for each Resource Provider (RP). This ensures equitable distribution without disproportionately affecting smaller GPUs.
📌 Why These Changes Matter
These updates optimize the IncentiveNet rewards system, ensuring that GPUs capable of running Lilypad Jobs (including AI jobs and Proof of Work) are rewarded fairly regardless of the size. GPUs of all sizes have a meaningful role in the network as they can run smaller workloads at the network edge. Naturally, high-performance GPUs will process more work and large workloads earning more rewards.. The refined calculation aligns with our commitment to fostering a fair and balanced ecosystem.
For a detailed breakdown of the new reward structure, check out our blog.
Stay up to date in the updates-rp Discord channel.
⚡ Developer Events & DX Updates
X/Twitter Space: Open Source & Decentralization
In our recent X/Twitter Space, we discussed the importance of open-source communities in driving decentralized innovation. The session covered strategies for fostering collaboration and growth.
Listen here: https://x.com/i/spaces/1vOxwrPkERoJB/peek
Weekly AMA Sessions
Curious about Lilypad’s latest developments? Join our Weekly AMA Sessions every Wednesday to:
Connect with the Lilypad team.
Ask questions and share feedback.
Stay informed on upcoming milestones.
📅 Flexible Timings: We alternate session times to accommodate our global community. Check the Lilypad Calendar for details.
🌟Community Spotlight
Our community continues to shine! Here are some recent highlights:
Ambassador Contributions: A special shoutout to Lilypad Ambassador Lead, Kabularasa for contributions to RP issue troubleshooting, building tools for the community, and even assisting in testing edge cases for Lilypad RPs!
Tadpole Talent Board: Looking for opportunities in AI or blockchain? Dive into our Discord job hub, where you can connect with collaborators and showcase your skills.
🌏 Lilypad at Devcon
Lilypad recently traveled to vibrant Bangkok for Devcon, Ethereum's premier conference for developers, thinkers, and innovators. It was an inspiring week filled with groundbreaking ideas, meaningful connections, and exciting side events that highlighted the future of decentralized technologies. We’re excited to bring these insights back as we continue building the future of accessible compute!
🔮 What's Next?
Updates to the IncentiveNet are in the pipeline with the goal of optimizing the job matching engine, improving the Lilypad implementation for RPs, and more. The Lilypad Whitepaper and Tokenomics are currently in review. A huge shout out to the Crypto Econ Lab for providing seasoned expertise when designing the Lilypad Mainnet economy.
Mainnet development continues with our engineering team running sprints on all parts of the network stack. We will continue to keep the community updated as the Mainnet development progresses.
Stay updated on all the latest developments in the updates-rp channel on Discord!
☎️ Contact Us
💬 Got questions?
Join our Weekly AMA Sessions hosted every Wednesday!
Chat with us on Discord or follow us on X/Twitter for real-time updates.
Explore the Lilypad Blog for insights into our projects and vision.
We’re here to support our community every step of the way! Together, we’re shaping the future of decentralized AI. 🚀