How does this participation take place?
When you install the application and launch it, you are launching your node and joining the network. The following steps happen under the hood:1
A query is submitted into the swarm network.This is the starting point where the entire decentralized network begins its work — there is a request in the network that requires a response.
2
Multiple nodes independently decide to engage with the query.Each node has unique expertise depending on the model that the node runs. This way, every node is capable in its own way, being able to handle certain types of queries better than others.Now, different nodes in the network (including yours) decide whether they will participate in processing specific queries that appear in the network. Each node acts autonomously, which increases the system’s resilience and reliability.To make it possible, your node loads an AI model into your VRAM or RAM depending on your system type. Your node also allocates resources for the KV Cache, which is essential for your node’s performance and determines what queries it can answer.
Currently, Fortytwo applications do not support full load balancing, instead KV Cache attempts to take up as much as 90% of a system’s free memory to ensure best possible performance of your node. We recommend stopping the app (for CLI — closing the terminal) before running resource-intensive applications (e.g. videogames, video editing, or graphics software) to prevent performance issues.This does not apply to nodes running on dedicated machines, where the node app runs independently without impacting other tasks.Future versions will include smart load balancing with editable cache, which will allow nodes to run alongside other resource-intensive applications without any noticeable performance impact.
3
Each participating node produces its unique response to the query.As your node is actively preparing its response to the query from the network, it increases resource consumption of your system for a very short time.As a result, the network receives multiple solution options created by participants with their unique knowledge.
4
Nodes compare their responses and assess their quality collectively.Your node evaluates responses made by other nodes that were processing the same query. Other nodes participating in that request also evaluate your node’s response in the same way.
5
The swarm converges on the most relevant responses.Collective comparison allows weaker options to be filtered out and stronger ones to be reinforced.
6
The result is sent back as the network’s collective answer to the query.This is where the most useful and accurate response gets defined out of all other responses. Then, the final merged result is sent back as the network’s collective answer to the query. This outcome reflects the combined judgement and verification of many independent query participants.
7
Rewards for the best responses.Nodes that provided the best answers are rewarded with tokens. The node that collectively was evaluated as the one that gave the best response gets the most of the reward.
To increase the chances of winning, it is important to choose the right model: Selecting a Model for Your Node.