Lifecycle of an AI Task

When a user publishes a task to the AegisAI network, the operational process is as follows

1) Submit an AI Task

To use the AegisAI network, users will first need to submit the task to the L1(The consensus layer) . This can be done by interacting with the L1 directly or through a cross chain message from other blockchains like Ethereum or Solana. After the submission is successful, the task information will be included into blocks of the L1 and open for resource nodes to complete it.

2) Computation Phase

Resource nodes will first pick tasks from the L1 and initiate the computation process. For diversity of the results, users can ask for multiple calculations with different random seeds. The first batch of nodes submitted required results to the L1 qualifies for a reward, with reward weights decreasing progressively based on ranking.

3) Verification Phase

Resource nodes verify the computation results already submitted to the L1, earning result verification rewards.

4) Soft Finality

Once a computation result receives a certain number of verifications without any conflicting results, it can be considered preliminarily confirmed. At this period, users can ask for a cross-chain message back to the original source chain so they can utilize the AI result.

5) Hard Finality

If no conflicting verification results emerge within 24 hours, the task result is finalized, rewards are distributed, and the task is officially concluded in the L1.

Balance the workload of resource nodes

To encourage resource nodes working on different tasks, there will be a privilege weight based on node address and task hash. This privilege weight will improve the rankings in reward distribution. So resource nodes will pick the tasks with a higher privilege weight when there are multiple uncompleted tasks in the network.