DeltaOne Explained

DeltaOne is the protocol's way of asking a strict question: did this contribution move the decision layer forward in a measurable way?

That framing matters because high-quality AI infrastructure cannot rely on vague engagement metrics. It needs a reward surface that maps to real changes in model behavior, benchmark quality, or production routing outcomes.

At an implementation level, contributions can be normalized into a scoring pipeline:

type Contribution = {
  baselineScore: number
  candidateScore: number
  confidence: number
}
 
export function computeDeltaOne(contribution: Contribution) {
  const improvement = contribution.candidateScore - contribution.baselineScore
  return improvement * contribution.confidence
}

The exact function will evolve, but the principle stays fixed: rewards should anchor to verifiable improvement, not narrative.