Skip to main content
A districuted AI network must ensure that every node is genuinely performing computation—not faking outputs to save electricity or GPU cycles. FAR AI solves this problem with a cryptographically enforced Proof of Compute layer, guaranteeing honest execution without requiring collateral staking. The “Lazy Node” Threat In open networks, malicious or cost-cutting nodes may attempt to:
  • Pretend to run the model while returning random or low-effort outputs
  • Shortcut inference by using smaller models internally
  • Replay old outputs instead of generating fresh responses
  • Drop computations entirely to save power
This type of behavior is known as lazy computation. If left unchecked, it causes:
  • Poor user experience
  • Lower model accuracy
  • Network instability
  • Unfair payouts to dishonest nodes
Because FAR AI pays nodes based on verified throughput, we must guarantee that every node actually computes what it claims. Probabilistic Proof of Compute (PPC) FAR AI uses a lightweight, probabilistic verification system that continuously tests nodes without slowing down real users. This system works like a cryptographic “lie detector.” How it works: 1. The Honey Pot Randomly (for example, 1 out of every 80–120 requests), the Orchestrator injects a Test Vector into the node’s workload.
  • It looks like a normal inference request
  • It behaves like a normal request
  • The node cannot detect that it is a test
  • The node must compute it exactly as it would any inference.
2. Ground Truth Known to Orchestrator Each Test Vector is generated from:
  • A known prompt
  • A known model state
  • A deterministic evaluation path
This allows the Orchestrator to know the correct answer in advance with extremely high certainty. 3. The Integrity Check When a node returns the output:
  • The system compares the node’s answer with the known solution
  • A slight tolerance window is allowed because of quantization variance
  • Deviations outside the threshold indicate the node did not compute the request faithfully
4. Automatic Penalties  Because nodes do not stake FAR, penalties use alternative mechanisms: A. Reputation Drop The node’s global reputation score decreases immediately. Reputation controls:
  • Routing priority
  • Eligibility for high-paying traffic
  • Participation in Prime Triads
  • Share of Npnts-based emission rewards
A node with low reputation becomes economically irrelevant. B. Reward Nullification For the cycle in which the cheating occurs:
  • The node receives 0 Npnts
  • Therefore earns 0 FAR rewards
  • Cheating becomes strictly unprofitable.
C. Automatic Ban If a node fails multiple Test Vectors:
  • The launcher bans it automatically
  • The node’s hardware fingerprint is blocked
  • The wallet address is added to the global denylist
  • Its prime partners (if any) are notified to recover gracefully
No human intervention required. Economic Security Without Staking Because node rewards scale directly with Npnts, and cheating eliminates Npnts, the system creates a simple equation: Honest computation > Cheating
Always.
Nodes maximize profits only by performing real work.