🌎 Overview

The Lilypad Network aims to democratize access to GPU compute for running AI models and other computationally intensive processes. By leveraging peer to peer networks and crypto-economic incentives, the Network aggregates idle compute capacity eventually providing alternative solutions to established Cloud providers (GPU services).

The network economy provides global access to idle computing accelerators (GPUs) in combination with developer tooling and software offerings. The types of processes that are run on the network vary widely, however, a primary stakeholder group that is served by this network economy is the AI/ML community. By opening up access to GPU compute, the Lilypad Network enables ML researchers and businesses running AI models to scale their systems using the Ethereum blockchain (Arbitrum L2) to verify compute jobs.

The service provided by this distributed network of GPUs is primarily architected for serverless tasks. Instead of renting a GPU for “x” amount of time, users pay per compute process. On the Lilypad Network, this is referred to as a Job. Resource Providers (RPs) earn a fee for every job run on their machine.

When evaluating the trade offs between privacy, verification, and performance (As illustrated by “Juan’s Triangle” below), the Lilypad team focused on serving end users and providing smooth onramps to leverage the benefits of distributed compute. The Lilypad Mainnet will use an optimistic approach with economic collateral in the form of staking.

The Lilypad economic model, approach to verifiability, approach to non-deterministic AI jobs, and more continue to be tested on the IncentiveNet! IncentiveNet is an incentivized testnet that has on-boarded over 1600 GPUs to date with thousands of jobs run since the inception.

The Resource Provider (RP) Leaderboard offers a comprehensive view of the network and Lilybit_ , further architecting the tech stack and reviewing the Lilypad Whitepaper. Mainnet has been inspired by learnings from IncentiveNet in addition to the deep expertise of our research team and CryptoEconLab. Work on rewards earned by each RP. As our team continues to build and refine this critical compute infrastructure, we invite you to run a node and join us on Discord in this exciting journey towards a more distributed and accessible AI ecosystem!

⚒️ Engineering Update

The last two weeks at Lilypad have been busy! The core engineering team continues to push towards Mainnet launch further architecting the tech stack and reviewing the Lilypad Whitepaper. Mainnet has been inspired by learnings from IncentiveNet in addition to the deep expertise of our research team working with CryptoEconLab. Work on the Network economics, Lilypad Tokenomics, verification mechanisms, privacy needs, and more continues! The Lilypad Whitepaper is expected in November 2024.

In addition to Mainnet work, the team continues to support the Lilypad IncentiveNet with improved slashing rules, more stable version releases for running jobs, and more! The IncentiveNet has provided a foundation for testing all Lilypad mechanisms in a production environment. Our team is proud to have scaled the network (Resource Providers) quickly over the last few months and continue pushing new partnerships and use cases with end users. From BioML to AI Agents, we are super excited to see what the community has been building on Lilypad!

RPs update to the latest version! - v2.7.0 Chongzhi Zu

“Zu's value of pi is precise to six decimal places, and for almost nine hundred years thereafter, no subsequent mathematician computed a value this precise.”

Lilypad v2.7.0 Chongzhi was released requiring Lilypad RPs on IncentiveNet to update! The latest Lilypad release is focused on ensuring stability for the network (the public RPCs + personal RPCs seems to be working well):

• RPC URLs provided by Lilypad have been deprecated in favor of the public RPC URLs added in v2.6.0.

• A Lilypad RP will check periodically to ensure Bacalhau v1.3.2 is running. If Bacalhau is not running the Resource Provider will turn off.

• Improved collection of machine specs from Bacalhau.

• Added Lilypad Solver tracing

Lilybit_ Rewards

The Proof of Work (PoW) mechanism used to verify RP uptime and hashrate is used to evaluate rewards earned by a RP. Each GPU generates a hashrate based on the size of the card. The hashrate and associated number of PoWs run per day are used to calculate the rewards credits given to RPs.

Each hour, Resource Providers will be tasked to run the proof and, additionally, randomly, many times per day. There is a hashrate call every four hours when the Proof of Work is run. RPs receive a multiplier relative to how many 4-hour windows of PoWs completed in that day. The graph below shows the daily multiplier for a given number of 4-hour windows of POWs completed across a 24 hour period.

Proof of Work (PoW) runs four-hour intervals per 24 hours.

An RP will lose the 4x daily multiplier if the hashrate call (every four hours) is missed. If the PoW happens in the window of time that a node is restarting/offline, then it's missed. In the case of RPs having downtime spanning multiple days, earning capacity will be slashed, decreasing their reward-earning potential. This is to encourage maximum GPU contribution and uptime. If you have further questions, feel free to ask in our #i-need-help channel on Discord.

🖥️ Tokenomics & IncentiveNet

With the support of CryptoEconLab, our research team continues to review and improve the Lilypad Whitepaper! Planning the network Tokenomics includes the Lilypad RP pricing system, job matching systems, ERC20 token emission schedule, and more. Our research team has been super thankful for the IncentiveNet RP community as many of these systems have been tested and discussed publicly. This process allows our team to stress test assumptions in the real world.

In addition to the Whitepaper, the Lilypad research team continues to support IncentiveNet. Based on community feedback, changes have been made to the network slashing rules providing some relief to RPs that are long term contributors (online for over 1 month).

Changes to slashing

The community has spoken and we have heard! Changes to the Lilypad slashing mechanism are currently undergoing testing with the first change released Monday Sept/29.

A grace period for RP downtime is now included in the slashing mechanism. RPs will earn 2 days of a “grace period” after every 30 days of continuous service provided. These 2 days will be applied to 2 subsequent down days recorded by the RP, allowing the RP to avoid slashing for these 2 days. Grace Period days do not accumulate to more than 2 days ever. Once used the 30 day count to obtain the 2 days restarts.

The next update to the slashing will include a tiered slashing % based on the number of days a RP is offline. The more consecutive days an RP is offline, the higher the Lilybit_ slashing %.

Stay connected on the Lilypad Discord and follow along on X!

⚡ Developer Events & DX Updates

Open Source rewards dashboard

Open source contributions to the Lilypad main repo as well as other public repos have been awesome! From unit testing to Discord bots, the community has stepped up to help grow the network. Our team is energized by the enthusiasm and as a result have decided to reward open source contributors with Lilybit_ rewards! Issues marked as “Community” will be the first contributions that are eligible for Lilybit_ rewards. Reach out to our team if you would like to propose an issue!

Check out the open source contributors dashboard to see rewards earned by the community!

This rewards initiative is part of a larger Community Rewards program that has been rolled out in phases. Initially, open source contributors will be eligible for rewards in addition to the Lilypad Ambassadors. In the next phase of Community Rewards, Lilybit_ will be available for creating educational content!

Weekly AMA Sessions

To help answer questions and gain insights from our team, Lilypad is hosting AMA sessions every Wednesday! We will alternate the timing of the AMA every week to ensure we can cater to RPs and devs across the globe! Find an AMA time that works for you on the Lilypad Calendar.

Catch up with the AMA from last week hosted by DeveloperSteve and Taosheng!

🚀 Partnerships

Uprising Labs Partnership

Our team is proud to partner with Uprising Labs to begin supporting compute jobs needed for gaming. This partnership represents a push into an additional vertical for the Lilypad network providing serverless execution for tasks within the Uprising Labs ecosystem.

Follow along on X and subscribe to our newsletter to stay up to date with announcements on this exciting partnership!

Expanding the Lilypad Ecosystem

Each partnership that has been announced or is in the pipeline helps to address critical components of the Lilypad ecosystem. Partner Resource Providers have joined the network, growing our enterprise grade GPU supply and building out integrations.

While these project collaborations are still in their early stages, we're actively exploring integrations that combine Lilypad's distributed computing power with our partners' unique capabilities. Here’s some of the integrations we’ve been partnering on.

• On-Chain Database & Storage: Leveraging IPFS and blockchain tech for enhanced security, transparency, and trust.

• Novel Web3 Applications (dApps): Creating new tools and experiences that wouldn't be possible without decentralized compute infrastructure.

🌟Community

Lilypad featured in Blockworks!

The Lilypad Network was featured in Blockworks with an opinion piece from our CEO, Ally Haire! Ally discusses problems faced by users and businesses with regards to existing GPU compute access. Read more to learn about how Lilypad solves problems with GPU accessibility and compute verifiability.

Pondside Chat with Edatweets_

In the latest Pondside Chat, Lilypad CIO Steve interviewed Eda Aktuk, Developer Advocate at Hyperlane. They explored Eda's journey from Web2 to Web3, her passion for education, and her experiences with ETHGlobal Hackathons and DAOs. The conversation examined the future of Web3, focusing on interoperability, modular blockchains, building organic communities and the role of DAOs in fostering innovation. Eda also shared her perspective on Lilypad's mission to democratize AI and its potential impact on the Web3 ecosystem.

Meme contest winners

Winners of the Meme-a-thon have been announced! Thank you to everyone who participated. A large influx of submissions at the last minute prompted our team to change the game plan! We chose our favorite memes and then posted a poll on X asking the community to vote. The vote was the final decision for the contest. Congratulations to @99Paiin who was voted winner with their very funny meme adaptation of a classic Mr Bean skit.

Lilypad Twitter Spaces - ‘Open Source: Community Driven Innovation’

The Lilypad team will be hosting an X Spaces every week to connect with the community. In case you missed it, check out the recording of the latest space where we discussed “Open Source: Community Driven Innovation”. Hear from the brilliant team at Lilypad as they discuss all things open source with guest appearance from dFusion.

This Space was perfect timing to align with our involvement in Hacktoberfest - check out the Lilypad main repo to get involved and look for the Hacktoberfest tag.

🔮 What's Next?

Contribute to the Lilypad Network during Hactoberfest! We have joined the open source fun and will be supporting contributions of all types.

Improvements to the IncentiveNet slashing mechanism, rewards calculation, and more are in the works. This work aims to improve the system economics as well as improve job reliability. As the network scaled quickly, our team ahs been carefully implementing changes to scale up the network infra accordingly.

Follow the updates-rp channel in the Lilypad Discord to keep up to date with the the latest RP updates and announcements!

☎️ Contact Us

💬 Have more questions or looking for guidance? Join our community AMA hosted every Wednesday!

Chat to us on Discord and follow along on X/Twitter for the latest news and updates!