Subscribe to wiki
Share wiki
Bookmark
Everlyn
The Agent Tokenization Platform (ATP):Build autonomous agents with the Agent Development Kit (ADK)
0%
Everlyn
Everlyn, also known as Everlyn AI, is an open-source, decentralized artificial intelligence (AI) protocol focused on the generation of video content. The project aims to provide a democratized alternative to proprietary, closed-source video AI models developed by large technology corporations. It operates as a decentralized physical infrastructure network (DePIN), coordinating a global network of GPU computing resources to facilitate video creation. The native utility and governance asset of the ecosystem is the LYN token. [1] [2]
Overview
Everlyn's stated mission is to create an "open dream machine" by developing foundational video models and a surrounding multimodal ecosystem that are accessible to a global community of users and developers. [3] The protocol is designed to address several key challenges in the generative AI space, including the centralization of powerful AI models, the high computational costs associated with video generation, and concerns regarding data privacy. [2] By leveraging a decentralized network of compute providers, Everlyn aims to reduce video generation costs by up to tenfold compared to centralized services. [4]
The project is developed by Lyn Labs, a technology company with a stated presence in Hong Kong, and operations across Asia, North America, and the Middle East. [4] [5] The founding team is comprised of a collective of anonymous and pseudonymous AI researchers, professors, and industry veterans. Team members reportedly have academic affiliations with institutions such as Cornell University, University of Oxford, Stanford University, and Peking University, as well as prior experience at major technology companies including Meta, Deepmind, Google, and Tencent. [6] [3] These contributors have been purportedly involved in the development of other significant AI video models like Google's Video Poet, Meta's Make-a-Video, and the open-source Open-Sora-Plan. [3]
History and Development
The Token Generation Event (TGE) for the LYN token occurred on February 28, 2024. [5] Following the launch, the token was listed on several centralized cryptocurrency exchanges in early March 2024. [2] [5]
According to the project's roadmap, a testnet for its "Creator Nodes" was planned for the second quarter of 2024, allowing compute providers to begin connecting to and testing the network. The launch of the Everlyn Mainnet V1 was projected for the fourth quarter of 2024. [5]
Technology
Everlyn’s technology is structured as a multi-layered ecosystem that combines a foundational AI model, a decentralized compute protocol, a user-facing platform, and autonomous AI agents. The platform claims to offer video generation speeds of around 15 seconds, a significant reduction from an industry average of several minutes, and at a cost up to 25 times lower than centralized competitors. [1]
Protocol Architecture
Everlyn operates as a decentralized network that coordinates tasks between three primary participants: users, GPU node providers (called Creator Nodes), and AI model creators. The workflow is designed to be permissionless and is managed by the protocol. [1] [5]
- A user submits a generative AI task, such as a text-to-video prompt, to the network. The user pays for this service using LYN tokens.
- The protocol routes the task to an available Creator Node on the network.
- The Creator Node utilizes its GPU computing power and an open-source AI model to process the request and render the video content.
- The network verifies the completed task.
- Upon successful verification, the Creator Node operator is compensated with LYN tokens from both the user's fee and protocol-based rewards. [5]
This model is intended to give users greater control and ownership over their data through on-device processing for certain tasks and encrypted data containers. The architecture is also designed for multi-chain interoperability to integrate with various Layer 1 and Layer 2 blockchains. [2]
Foundational Model: Everlyn-1
The core of the ecosystem is Everlyn-1 (also referred to as Openlyn), described as an open-source, autoregressive foundational video model. [2] [7] The model operates through an autoregressive process, generating video frame by frame in a manner similar to how large language models predict the next word in a sentence. This approach is designed to enable real-time, continuous video generation of potentially unlimited length. [2]
The stated research objectives for the model include improving output duration, enhancing interactivity with generated content, reducing generation latency, and achieving hyper-realistic visual quality. [6] To achieve these goals, the development team is focused on several technical innovations:
- Efficiency Enhancements: Techniques such as Superpipeline-Quantization and xDiT are used to improve efficiency and lower computational requirements. [2]
- Data Processing: The project utilizes a proprietary data pipeline designed to accelerate the video clipping process for model training. [2]
- Video Editing: A framework called Composite Video Generation (COVG) enables precise video editing by integrating text and video conditions. [2]
- Hallucination Reduction: A MultiModal Robustness (MMR) benchmark is used to test and minimize the model's generation of incorrect or nonsensical content. [2]
- On-Chain Verifiability: Each AI-generated video can be anchored to blockchain metadata, creating a verifiable record of its creation, provenance, and authenticity. [4]
Everworld and Autonomous Video Agents
Everworld is the user-facing video creation engine and platform built upon the Everlyn-1 model. It serves as the interface for users to create and interact with photorealistic, AI-driven digital representations, which are called Lyns or autonomous video agents. [3] [6]
These video agents are designed to be more than just avatars; they are envisioned as functional, task-oriented digital proxies for users. Each agent possesses its own on-chain wallet, enabling it to perform tasks, execute financial transactions, manage assets, and interact with other agents and decentralized applications (dApps) independently. [2] A fixed, one-time fee paid in LYN is required to mint a new video agent. [2]
Potential use cases for these agents include:
- Personal Automation: Managing schedules, ordering food, booking travel, or calling rideshares.
- Professional Tasks: Attending virtual meetings, delivering presentations, or handling customer service inquiries.
- Social & Entertainment: Managing social media accounts, screening matches on dating applications, or creating personalized gaming experiences.
- Agent-to-Agent Economy: Interacting and transacting with other agents in a decentralized marketplace. [2]
Agent API (AAPI) Ecosystem
The Agent API (AAPI) ecosystem functions as a decentralized marketplace where developers can build, deploy, and monetize new skills and functionalities for the video agents. Developers can offer these new capabilities through LYN-based subscriptions. LYN token holders can stake their tokens on specific AAPIs they believe will be successful, allowing them to earn a share of the revenue generated by that API. This creates an economic incentive for the development of a diverse range of agent abilities. [2]
Partners and Advisors
Prominent AI researcher Professor Yann LeCun is listed as an advisor to the Everlyn project. The project's infrastructure partners include the decentralized GPU marketplace Aethir, as well as the BNB Chain and Sui blockchains. [4]
LYN Token
The LYN token is the native asset of the Everlyn ecosystem, serving roles in payments, staking, governance, and incentivization. The token exists on multiple blockchains, including the BNB Smart Chain as a BEP-20 token and the Ethereum blockchain as an ERC-20 token. [4] [1]
Tokenomics
Everlyn launched with a genesis total supply of 1,000,000,000 LYN. The tokenomics model includes a programmed inflation mechanism to fund ongoing network rewards. This inflation starts at an annual rate of 10%, with the rate decreasing by 10% of the previous year's rate each subsequent year. This model continues until the inflation rate stabilizes at a perpetual issuance of 1% per annum. [2]
The initial allocation of the 1 billion LYN genesis supply was distributed as follows:
- Node Rewards Emission: 20.00% (200,000,000 LYN)
- R&D & Ecosystem: 23.60% (236,000,000 LYN)
- Initial Core Contributors: 19.80% (198,000,000 LYN)
- Public Allocation: 15.00% (150,000,000 LYN)
- Early Backers (Pre-Seed & Seed): 14.70% (147,000,000 LYN)
- Early Backers (Series A): 6.90% (69,000,000 LYN)
The public allocation was fully unlocked at the Token Generation Event. Allocations for the team, investors, and other categories are subject to various vesting schedules, withsome lasting up to three years. [2] [4]
Utility
The LYN token is integral to the functioning of the Everlyn ecosystem. Its primary utilities include:
- Payments: LYN is the exclusive currency for paying for video generation services and other network tasks. [1]
- Agent Creation & Operation: A one-time fee in LYN is required to mint a new video agent, and LYN is used to pay for gas fees for all on-chain actions performed by agents. [2]
- Staking: GPU providers must stake LYN to operate a Creator Node, which acts as an economic bond to ensure honest behavior and service quality. Staking LYN in an agent's wallet can also unlock premium features. [2] [5]
- AAPI Marketplace: All transactions within the Agent API marketplace, including subscriptions, developer earnings, and staking rewards, are conducted in LYN. [2]
- Governance: LYN holders can participate in the project's governance by voting on proposals related to protocol upgrades, resource allocation, and other key decisions through the Lyn DAO. [2]
- Incentives & Revenue Sharing: The token is used to reward network participants, and holders can stake LYN on successful agents or AAPIs to earn a share of the revenue they generate. [4] [2]
Contract Addresses
See something wrong?
The Agent Tokenization Platform (ATP):Build autonomous agents with the Agent Development Kit (ADK)