Gradual Consensus
Act quickly, react faster, and reward slowly.
While the task standard supports a wide range of possible applications, the most straightforward approach is to follow a multi-epoch confirmation process, which we refer to as Gradual Consensus.
Why is it gradual?
One of the major limitations of stake-based games is the potential for an actor with a large amount of tokens to temporarily take over the system and potentially even extract rewards unfairly. Koii Tasks naturally avoid this issue by slowing down reward distributions and using proportionately high numbers of audit nodes compared to task executors.
How does it work?
Gradual consensus is a multi-epoch game under which nodes first perform a task, and then vote on a reward distribution. It's important to understand that while the process takes two epochs to complete, each cycle operates independently once per epoch, such that there is always one cycle starting and completing within every epoch.
Epoch 1:
- Task nodes run the task executable and write new records on-chain.
- Audit nodes verify the previous epoch's contributions and vote to slash stake
Epoch 2:
- Task nodes evaluate and aggregate each other's submissions and vote to confirm
- Audit nodes verify the previous epoch's contributions and vote to slash stake
At first, it can be difficult to imagine this process, but a helpful analogy would be a batch production line, where one group produces a product all day while the other evaluates its quality and may reject entire batches or penalize individual workers for sub-par performance.
Applications
Gradual Consensus works well on tasks that have a deterministic outcome but requires some significant computation that cannot be done on-chain on networks like Ethereum or Solana due to computation costs or network call limitations. Some particular applications include web scraping, indexing on-chain content, compiling or hosting web apps or sites, or even managing databases and caches of on-chain content.