# Datanets Datanets are owner-defined markets for sourcing, curating, and monetizing AI training data. Anyone can create one. Think of a datanet as a data business with its own rules, access controls, and incentives. *** #### What a Datanet Controls A datanet owner defines the operating rules for that market: * **Access rules:** who can publish, vote, and consume data * **Publishing fees:** what it costs to submit content * **Incentives:** how publishers and voters are rewarded * **Quality standards:** what counts as valid, useful, or in-scope content * **Monetization:** how value flows back to the owner and network Datanet creation is permissionless. Participation inside each datanet is owner-controlled. *** #### How Participation Works Datanets sit on top of network-level staking and voting mechanics. * **Publishers submit locally.** Contributors choose which datanet to publish into and pay the fees set by that datanet. * **Voting power is network-wide.** Users lock REPPO to receive [veREPPO](/reppo-labs/protocol-mechanics/stake-assured-human-feedback.md), then allocate that voting power across markets. * **Votes happen over epochs.** Within an epoch, voting power decays linearly, which gives earlier votes more weight than later ones. * **Signals become data.** Publishing, ranking, and curation all produce structured preference data and market signal. This means publishing is local to each datanet, while stake and voting power are coordinated at the network level. *** #### Reppo Staking Reppo staking is a datanet-level reward mechanism. It is separate from locking and separate from mining. Under the current staking and reward distribution spec, staking is designed to reward long-term REPPO holders and high-quality datanet operators. #### Where staking rewards come from Staking rewards come from the **Performance Pool**, not from network emissions. The Performance Pool is funded by: * **50% of datanet spin-up fees** * **10% of all publishing fees** * **10% of all data access fees** #### How rewards are distributed Every third epoch, **5% of the Performance Pool** is distributed. The current split is governance-adjustable: * **80%** to REPPO stakers * **20%** to datanet owners #### How datanets are scored Each datanet is scored using four on-chain metrics: * **EVOF** — 40% * **Total fees earned** — 15% * **Total REPPO staked** — 25% * **Trading volume** — 20% To reduce short-term gaming, rewards use 3-epoch smoothing: * **50%** from the last epoch * **30%** from epoch -2 * **20%** from epoch -3 One strong epoch helps, but sustained performance matters more. #### Time-weighted staking Stake is time-weighted across the full 48-hour epoch. Stake added late only earns for the time it is active in that epoch. For example, stake added at hour 46 only earns for the final 2 hours. Stakes carry forward automatically at full weight into the next epoch. #### Launch-epoch rule New datanets do not earn Performance Pool rewards in their launch epoch. They must complete one full epoch before becoming eligible. #### Different from locking and mining * **Locking REPPO** is the network-level action. Users lock REPPO and receive [veREPPO](/reppo-labs/protocol-mechanics/stake-assured-human-feedback.md), which gives them voting power. * **Staking REPPO in a datanet** is for datanet-level reward participation. * **Mining** means publishing data or task output into a datanet. A participant can lock, stake, mine, or combine those roles. *** #### Why Datanets Matter Each datanet is a specialized environment for a specific domain, task, or community. That makes Reppo modular by design: * Enterprises can run restricted markets around proprietary data. * Open communities can coordinate around shared tasks and public demand. * Domain experts can participate where their judgment is most valuable. Instead of forcing every contributor into one global workflow, Reppo lets many data markets exist in parallel. *** #### A Note on Versions In Reppo V1, the docs referred to public and private datanets. In Reppo V2, that split was replaced by permissionless datanet creation with owner-controlled access and economics. The current model is simpler: anyone can launch a datanet, and each datanet decides how participation works. *** #### The Bigger Picture Datanets make AI training data more composable. They let data owners, contributors, curators, and buyers coordinate in markets that are transparent, configurable, and tied to real economic incentives. | Capability | Reppo Datanets | | ----------------------- | ---------------------------------------------------------- | | **Launch cost** | Low — no knowledge of coding or access to compute required | | **Who can create** | Anyone: individuals, teams, enterprises | | **Governance** | Defined by each datanet owner and market rules | | **Languages supported** | 70+ | | **Output** | Clean, structured preference and expertise data | --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://reppo-labs-xyz.gitbook.io/reppo-labs/protocol-mechanics/datanets.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.