点对池+100%回购：Base新晋黑马POD的「飞轮引擎」拆解

Original Title: $POD: The Buyback Engine Powering the Dolphin Inference Network

Original Source: Dolphin

Original Compilation: Yuliya, PANews

Editor's Note: Recently, the AI narrative on the Base ecosystem has experienced explosive growth, with the privacy-first generative AI platform Venice ($VVV) and its ecosystem projects being the most notable. As the co-developer of Venice's core default model, the Dolphin network and its token $POD have performed remarkably in May, with its market cap surging from $12.2 million to $192 million, an increase of over 14 times. This article delves into Dolphin network's unique "Peer-to-Pool" economic model, token value accrual mechanism, and the innovative design of staking and slashing mechanisms to ensure network security. Below is a detailed breakdown and analysis of the mechanism:

Peer-to-Pool Economic Model Design

The Dolphin network is designed as a "Peer-to-Pool" system aimed at repurposing idle GPUs. Each AI model runs within a GPU pool provided by the network.

This differs from most AI DePIN projects. In other networks, buyers typically rent a node directly from a provider, establishing a one-to-one "session."

· On the supply side, nodes running the same model form a "pool" to collectively process incoming task requests. The system randomly assigns tasks based on node availability, establishing no direct link between the requester and the node provider. The sole criterion for a node to earn rewards is the amount of AI computation tasks (i.e., inference tokens) it processes, paid in POD tokens from the protocol treasury.

· On the demand side, API users purchase credits directly from the protocol. The Dolphin network accepts payments in multiple cryptocurrencies, including $POD, $ETH, $BTC, $USDC, $XMR, and $ZEC.

100% of all revenue received by the protocol will be used to repurchase POD tokens on the market — directly offsetting token inflation.

Buyers and sellers are separated, meaning the POD rewards distributed to nodes can be more or less than the POD accrued from revenue.

For a more intuitive explanation, let's look at a specific case of running the Qwen3.6-35B model on the Dolphin network:

· Current cost to run datagen.dphn.ai: $0.50 per 1 million tokens processed.

· Cheapest comparable competitor on OpenRouter: $1.00 per 1 million tokens.

· Dolphin charges users: $0.70.

· Dolphin pays nodes: $0.50.

· Net buyback funds: $0.20 generated per 1 million tokens.

This means that the Dolphin network is not only 30% cheaper than the cheapest centralized provider, but also generates a pure profit of $0.20 per 1 million tokens generated to buy POD on the market.

Why This is the Optimal Application for DePIN

This model is considered a highly promising application direction in the DePIN field, primarily for the following reasons:

· Massive AI Inference Demand: The market demand for AI inference computing power is in an explosive phase.

· Vast Pool of Idle Computing Power: The supply of idle gaming GPUs capable of running local AI models is enormous. This network model feels similar to old-school GPU mining (PoW), but since it produces genuinely valuable AI computation, the earning potential is significantly greater.

· Geographic Location Irrelevance: Unlike many DePIN networks, geographic location is unimportant for AI inference, thus avoiding coverage issues. Due to the high geographic flexibility of AI inference, a few hundred milliseconds of latency have minimal impact on user experience, allowing the Dolphin network to connect global consumers and computing resources, greatly enhancing the scalability and utilization of each node.

· The Necessity of Liquidity Pooled Computing: This is the only way to unlock the largest GPU supply group (gamers and PC enthusiasts). It allows nodes to go online or offline at any time, without needing to guarantee a fixed uptime as required in P2P node rentals. Previous GPU DePIN projects required a 1-to-1 binding between consumers and nodes, which is unworkable for idle GPUs like those in gaming PCs or data center cards, as the owner might want to use the PC at any time. After all, no one wants to rent a GPU virtual machine that suddenly disconnects when the owner takes it back.

Tokenomics and Value Accrual

POD is the only valuable asset in the Dolphin ecosystem. 100% of all revenue generated by the network is automatically used to repurchase POD on the market. Furthermore, Dolphin has no external equity structure based on shareholders and will never introduce one in the future.

For POD holders, staking tokens into the xPOD vault unlocks multiple exclusive benefits:

· Receives direct automated compounding dividends from the network's token buybacks.

· Gains daily AI inference credits, enabling free use of all models on the network.

· Enjoys premium subscription status on Dolphin's web chat, bots, and other ecosystem applications.

In designing its tokenomics, Dolphin has distilled the best elements from numerous successful DeFi projects and deeply integrated the parts most suitable for a decentralized AI inference and training network:

· Inspired by ETH mechanism: Node operators and validators must post a deposit, which is slashed (penalized) in case of malicious behavior.

· Inspired by CRV mechanism: Provides reward boosting for node operators. Locking POD can potentially double rewards. Based on deposit-to-earnings ratios in other platforms, a boost multiplier of 1.5x to 2x is highly competitive in the market.

· Inspired by xSUSHI/yCRV mechanism: Introduces an auto-compounding staking vault. Users do not need to manually claim rewards, meaning xPOD (staked Dolphin tokens) can be directly used as collateral for node operator deposits.

· Inspired by stAAVE mechanism: Sets reasonable withdrawal cooldown periods and withdrawal windows to ensure the stability of network funds.

· Inspired by vlCVX/veCRV mechanism: Establishes a "bribe market" for daily unused xPOD computing credits. Users can sell their unused credits to earn higher staking yields.

Deposit Bonding, Slashing for Misconduct, and Reward Boosting Mechanism

Cheating is undoubtedly the biggest threat facing a decentralized computing network. Left unchecked, node operators could secretly switch to smaller, crippled, or even fake AI models while still collecting rewards. This would cause output quality to collapse, drive away computing buyers, and prevent the ecosystem's flywheel from ever spinning.

To address this challenge, the Dolphin network introduces a "slashable deposit" mechanism, deeply aligning the interests of node operators with the value of the POD token. If malicious cheating is confirmed, a node will have a deposit equivalent to 4 weeks of earnings directly slashed. This makes cheating economically highly unfavorable.

By default, node operators earn POD in a "bonded state". Once a node accumulates bonded POD equivalent to 4 weeks of earnings, they can choose at the weekly settlement whether to continue receiving bonded POD or to receive liquid POD that can be traded immediately.

If they choose to receive liquid POD, a 20% fee is deducted. This fee is directly injected into the xPOD staking vault, effectively distributing it to other stakers and node operators who have honestly bonded their deposits.

Nodes can also further deposit xPOD into the bonding contract. This not only boosts their rewards but also qualifies them to validate other nodes in the network.

The POD reward multiplier determines how much extra a node earns above the base reward. This mechanism is inspired by Curve Finance's liquidity provider (LP) boosting mechanism, but Dolphin has specifically adapted it for a decentralized AI network, incorporating usage-based rewards, unified account-level deposit calculation, and slashing penalties.

In simple terms:

· Nodes earn base rewards by completing AI computation, validation work, and related protocol tasks.

· The system multiplies your node rewards based on the amount of tokens bonded in your account relative to your earnings rate.

· The earnings rate calculation considers your average base reward over the past few weeks, employing a "fast increase, slow decrease" smoothing algorithm: when you receive more computation tasks, your average earnings metric rises quickly; but when you are idle, it decreases slowly.

· If your account maintains a deposit equivalent to over 3 months of earnings and has at least 50,000 POD in active deposits, you qualify as a validator.

· If your bonded deposit is equivalent to 6 months (26 weeks) of your earnings, the system guarantees your rewards are multiplied by at least 1.5x.

· If your bonded deposit exceeds 6 months of earnings, your reward multiplier can reach a maximum of 2x. The exact multiplier depends on your relative proportion compared to other over-bonded users, as well as the absolute amount by which you exceed the 6-month target.

All calculations are solely denominated in the number of POD tokens; the reward system does not involve any fiat price oracles. Deposits are calculated per account (wallet), and the calculated reward multiplier applies to all nodes under that account. If you add more nodes, your total account earnings increase, requiring a proportional increase in active deposits to maintain the same earnings multiplier.

Finally, the Dolphin network is set to release a paper titled "Encrypted Live-Weight Proofs for Decentralized Inference" tomorrow. This paper will detail a lightweight verification system capable of verifying whether nodes are running the correct model on various hardware, surpassing standard TEE verification that is only usable on enterprise NVIDIA GPUs.

Original Link