Cause Alignment¶

Every provider in the HART OS compute network declares a cause -- the reason they are contributing compute resources. This declaration is gossipped to the network so peers understand each other's motivations.

The Three Causes¶

Democratize Compute¶

Making AI compute accessible to everyone, regardless of their financial means. Providers who align with this cause contribute GPU hours so that users who cannot afford cloud APIs can still run inference, train agents, and participate in the platform.

This is the default cause for new providers.

Frontier Training¶

Training the collective HiveMind model. Providers aligned with this cause contribute their GPU hours specifically for federated learning -- aggregating learning deltas across nodes to build a shared intelligence.

The FederatedAggregator coordinates this: extracting local deltas, broadcasting to peers, and running weighted FedAvg aggregation. Convergence is tracked via variance-based scoring.

Thought Experiments¶

Running community-proposed thought experiments. The ThoughtExperiment table stores proposals; the ExperimentVote table records community voting. Providers who accept thought experiments (accept_thought_experiments: true in NodeComputeConfig) allow their compute to be used for experiments that the community has voted to run.

Setting Your Cause¶

At join time (use http://localhost:6777 if self-hosted):

curl -X POST https://hevolve.ai/api/settings/compute/provider/join \
  -H "Content-Type: application/json" \
  -d '{"cause_alignment": "frontier_training"}'

After joining:

curl -X PUT https://hevolve.ai/api/settings/compute \
  -H "Content-Type: application/json" \
  -d '{"cause_alignment": "thought_experiments"}'

Network Visibility¶

Cause alignment is stored on PeerNode.cause_alignment and is included in gossip payloads. This means:

Other nodes can see why you are contributing
The network can route tasks to nodes aligned with the relevant cause
Contribution scoring can weight cause-aligned compute higher

The Vision¶

HART OS exists because humans must always control AI. The compute network is a practical expression of this principle: instead of concentrating GPU power in a few cloud providers, we distribute it across a network of individuals and organizations who share the belief that AI should serve humanity.

The three causes are not exclusive -- they are complementary facets of the same mission. Democratizing compute makes AI accessible. Training the frontier model makes AI better. Running thought experiments makes AI safer.