Welcome to Reppo Labs!
Last updated
Last updated
In his most recent talk, Ilya Sutskever acknowledged that data is the biggest blocker for AI progress and synthetic data isn't cutting it. Reppo's north star is to unlock access to proprietary data locked in walled gardens off-chain to constantly provide fuel needed to improve AI systems. Years of experience at Filecoin made us realize that proprietary data remains locked because of three intertwined reasons - Security/privacy concerns, lack of perpetual incentives related to value generated from data assets, and lack of optimal price discovery mechanisms for niche datasets. Solving for one just doesn't cut it.
To address this, we are building the world's first Intent-Centric Data Exchange that enables AI Devs and Agents to discover, negotiate, commit, and access niche and proprietary datasets in real-time while aiding price discovery for datasets. Instead of earning a one-time access fee, data owners are are incentivized through IP co-ownership in the derived AI assets from their data. Datasets can be pooled, and pool owners on Reppo monetize their assets. Data is secured by Gateway's Encrypted Data Vaults, and is accessed by independent developers and agents who leverage it to build specialized models, agents, and applications i.e. AI assets. These AI assets are further monetized independently by their developers and owners in different markets with adaptable business models, are used as alternative data sources for DeFAI and TradFi use-cases, as well as inputs for training data for larger foundation models using multi-linear operators.
In addition to Data, we are also experimenting with coordinating other AI resources like Capital and Infra using Intent-Centricity and IP Co-Ownership. More on that later.
This is how it works:
On Reppo, Independent developers and/or AI agents i.e. Builders are organize into Pods. Humans and AI Agents collaborate on AI models, agents, and applications approach the community of Reppo token holders directly to earn their vote of confidence. Instead of token holders directly funding the builders, they lock Reppo tokens to get VeReppo tokens and redirect protocol emissions to the most-promising projects. The longer they keep their ownership stake in the Pod, the higher fees they accrue as a share of the fee generated by the pods.
While builders can bring their own data and AI assets in the pod, they have in platform access to the Reppo Data Exchange which leverages Anoma's intent-centric architecture to facilitate demand and supply for unique niche data, effectively helping with price discovery of niche datasets and creating monetary +IP co-ownership incentives for data owners who otherwise would not have shared their data assets for secure off-prem compute jobs.
Builders will also have ready access to all AI Infra they will ever need to bring their ideas to life, right within the Reppo Platform.
Reppo Protocol powers this entire economy. As the data is never sold, data owners become co-owners in the AI assets pods build. Additionally, anyone who actively contributes to pods beyond the investment mechanism discussed above can contribute their time and skills, as defined by the requirements of the Pod owner, to get a share of IP Co-ownership in the Pod's AI Assets.
The protocol brings together various stakeholders together:
Context Owners- Anyone with domain-expertise who has an idea but not the time or skills needed to build the AI model/agent/application.
Data Owners - Anyone with niche dataset(s) who doesn't have demand for the data or does not want to sell the data but still wants to monetize/benefit from it
Reppo Builders create and deploy AI model/agent/applications, referred to as AI assets. They might choose to collaborate with context and data owners in exchange through IP Co-ownership or they might bring their own context and data, only seeking protocol funding to bootstrap their idea. When AI assets perform well i.e. generate fees from usage, they earn rewards and share of their revenue generated is redistributed to those who "backed" their pod. Building a Pod is permissionless and cost a small fee, from experienced AI engineers to a tinkerer.
VeToken Voters provide the economic foundation for the game by staking $Reppo. They not only earn yield on their Reppo but also revenue from fees, share of protocol fees, and in some cases, revenue from data licensing.
On the Reppo Network, there is no subjective "Quality assessment". Rather, it happens in real-time through tracking votes, usage of AI assets, and historical track record i.e. Reputation. The function determining the share of fees shared between various stakeholders is governed through the Reppo Protocol and can be adjust dynamically based on overall ecosystem performance, ensuring continuous improvement.
The protocol creates value for multiple stakeholders:
For Builders, it offers an opportunity to bootstrap without VC Capital. The competition format provides immediate feedback on their ideas and PoCs, helping builders improve their AI assets over time. As assets perform better, builders can scale their earnings by deploying multiple assets in their pod, buying off IP stake from other stakeholders or entering even higher-stake competitions.
For VeToken Voter, it provides a way to earn yield and revenue while supporting AI innovation. The staking mechanism is incentive aligned for long term sustainability with multiple revenue streams ensuring consistent returns. Stakers can choose their level of participation, from voting to IP Co-ownership and active involvement in protocol and network governance.
For the AI Ecosystem, it generates a continuous stream of specialized models and agents that can be consumed as alternative data for quantitative finance or training data for AI and multi-agent systems.
The competitive format creates explicit demand for niche datasets, ensuring only the best outputs are rewarded, propelling infinite sum games in the field of AI.
On Reppo Network, the AI IP is verifiable by anyone, with provenance enforced at each stage of the production lifecycle including fingerprinting data sources, storage and compute providers, data transformations etc. Builders on Reppo also have on-demand access to AI Infrastructure through the ModelRivet engine, where they can merely express their intents and access infra optimized for their use case including but not limited to elastic storage and compute, proof attestation services, out-of-box verifiable inference endpoints etc.
Reppo's increases access to both existing infrastructure as well as other collaborators by incentivizing and automating enforcement of ownership, control, and monetization of AI assets i.e. models and agents through technological means. Once deployed on the Reppo Protocol, these assets can be composed infinitely in a permissionless manner while enabling value capture and revenue sharing with data owners and community contributors so that developers can lessen their dependency on hackathon prize money and grants and for the first time, become true owners of the IP they help create.
We believe on-chain data alone is an incomplete to build robust and high performing AI, especially for use cases like DePINs where utility is generated off-chain. We solve this using a network of network appraoch of context-specific Reppo Intelligence Pods, taking a optimistic approach that attaches ownership and reputation to generative and predictive AI built by pod owners. Each pod has individual ownership and monetization pillars, and pod owners actively govern these pillars in accordance with their values and priorities. The protocol, platform, and tools we are building help pod owners accelerate their efforts.
We are driven by making community-led and funded AI entrepreneurship the obvious choice for independent developers as their experimentation and innovation playground.
