Key Points

Lagrange is a decentralized infrastructure platform consisting of three main components: the ZK Prover network, the ZK co-processor, and DeepProve, the zkML library.

The project allows developers to perform complex computations off-chain and prove the validity of results using zero-knowledge proofs (ZKP).

With Lagrange, you can verify historical data, perform off-chain computations, and securely transfer information between networks.

Introduction

Imagine you want to calculate the average price of ETH across several blockchains. This is not an easy task, especially if you don't want to use oracles or pay high fees for access to historical data.

Lagrange simplifies the process by moving complex computations off-chain, generating zero-knowledge proofs (ZKP), and verifying results on the blockchain. This allows for computations, data movement, and confirmation across various blockchains to be done more safely and efficiently.

What is Lagrange

Lagrange is a cryptocurrency infrastructure project consisting of three main products: the ZK Prover network, the ZK co-processor, and DeepProve, the zkML library.

ZK Prover Network

The Prover from Lagrange is a decentralized network of operators capable of generating ZKP on demand. If a decentralized application (DApp) needs to prove the correctness of computations, it simply sends a request to Lagrange.

The Prover network performs complex mathematical computations off-chain and returns a compact proof that can be verified by smart contracts. The DApp does not need to recompute on the blockchain or involve a third party for verification.

Unlike traditional schemes that use a single coordinator (which can become a vulnerability), Lagrange divides its network into independent subnetworks. This means that multiple blockchains, rollups, or applications can use the platform simultaneously, and the network itself scales with increasing load.

ZK Co-Processor

The ZK co-processor allows querying data from the blockchain without requiring trust. Developers can write SQL queries to extract data from the smart contract storage across thousands of blocks, perform computations (e.g., averages or sums), and obtain a ZK proof. This proof can then be directly embedded in a smart contract and verified.

This feature works across different blockchain networks, allowing data to be queried from a layer two network, such as Base, and verifying the result on Ethereum without using a bridge. Now developers have another option for querying blockchain data — in addition to creating their own indexers, writing complex logic, or using centralized APIs.

zkML DeepProve

DeepProve is Lagrange's zkML system that allows developers to prove that their artificial intelligence (AI) models are functioning correctly without revealing the model itself or its inputs. It generates ZKP for the outputs made by the AI, allowing anyone to verify that the prediction was generated using the correct model and was not altered. This makes AI more transparent and reliable.

How Lagrange Works

Prover Network

The Prover network operates on the EigenLayer infrastructure and is supported by over 85 institutional-level AVS operators. These operators run a lightweight program (worker binary) that tracks incoming tasks. When a developer needs a proof, the network sends a task, and the designated operators perform the computations off-chain before returning the ZKP.

It is important for operators to be reliable: if they do not complete tasks correctly and on time, they risk losing part of their staked funds. The Prover network supports different verification systems, such as Plonky2 and Plonky3, and in the future, others depending on needs.

DARA (Double Auction Resource Allocation)

DARA is a resource allocation system based on Lagrange that powers the Prover network. The system matches developers who need computations with operators willing to perform them. Here's how it works:

The developer determines how much computation is needed and how much they are willing to pay for it.

The operator specifies how much power they can offer and their costs.

DARA automatically matches both sides, ensuring that developers do not overpay and operators receive fair compensation.

The system is designed to ensure efficiency and fairness: payment is only charged when the entire workload can be completed. If the network cannot fully process the request, it will not be executed. By encouraging honest bidding and excluding partial or fabricated results, DARA helps establish a functioning market.