@a0n/aeon-bazaar

Negotiation engine powered by void walking. Unbounded -- the market runs forever, no termination guarantee, no mediator. Agents interact directly on gnosis topologies.

For bounded dispute resolution with convergence certificates, see aeon-neutral.

Install

bun add @a0n/aeon-bazaar

Quick Start

import {
  createVoidBoundary,
  updateVoidBoundary,
  complementDistribution,
  createMetaCogState,
  c0Choose,
  c0Update,
  c1Measure,
  c3Adapt,
} from '@a0n/aeon-bazaar';

// Create an agent with 3 choices
const boundary = createVoidBoundary(3);
const state = createMetaCogState(3);

// Choose, observe outcome, update void boundary
const choice = c0Choose(state);
const outcome = { accepted: false, reward: -1 };

const updated = updateVoidBoundary(boundary, choice, outcome);
const complement = complementDistribution(updated);

// Metacognitive adaptation
const measurement = c1Measure(state, updated);
const adapted = c3Adapt(state, measurement);

bazaar vs neutral

| | aeon-bazaar | aeon-neutral | |--|------------|-------------| | Mode | Unbounded negotiation | Bounded dispute resolution | | Termination | No guarantee | Convergence certificate | | Mediator | None (walkers interact directly) | Third walker on the convergence site | | Use case | Markets, open-ended bargaining | Mediation, arbitration, settlement | | Walkers | 1+ self-interested agents | 2 game walkers + 1 Skyrms site walker |

API

Void Walker (core primitives)

| Export | Description | |--------|-------------| | createVoidBoundary(n) | Initialize void boundary for n choices | | updateVoidBoundary(boundary, choice, outcome) | Record outcome, grow the void | | decayVoidBoundary(boundary, rate) | Time-decay old observations | | complementDistribution(boundary) | Compute distribution from what didn't work |

Measurement

| Export | Description | |--------|-------------| | shannonEntropy(dist) | Information content of a distribution | | excessKurtosis(dist) | Peakedness -- how concentrated is the distribution | | inverseBule(dist) | Topological deficit measure | | giniCoefficient(dist) | Inequality measure across choices |

Metacognitive Walker (c0-c3)

| Export | Description | |--------|-------------| | createMetaCogState(n) | Initialize walker state for n choices | | c0Choose(state) | Choose action from complement distribution | | c0Update(state, choice, outcome) | Update walker after observing outcome | | c1Measure(state, boundary) | Measure kurtosis, entropy, inverse Bule | | c2SelectGait(state) | Select trot/canter/gallop gait | | c3Adapt(state, measurement) | Adapt exploration/exploitation balance |

Gait System

Three gaits for adaptive exploration:

| Gait | Depth | Behavior | |------|-------|----------| | Trot | Shallow | Broad exploration, high entropy | | Canter | Medium | Balanced exploration/exploitation | | Gallop | Deep | Focused exploitation, low entropy |

GAIT_DEPTH maps each gait to its pipeline depth for the laminar kurtosis pipeline.

Performance

• 1,548 rounds/ms (laminar pipeline, depth 16)
• 2,084 rounds/ms (adaptive gait)
• 10,000 negotiations in 6.5ms
• 17.3x throughput at pipeline depth 16

The Skyrms Equilibrium

The fixed point of the void gradient flow. Self-interested agents with perfect knowledge of the void boundary converge to strategies 55 percentage points more cooperative than Nash in Hawk-Dove.

Named for Brian Skyrms (Evolution of the Social Contract, 1996).

Development

bun test

Related

• aeon-neutral -- bounded mediation with convergence certificates
• Gnosis -- GGL language for negotiation topologies
• Aeon -- TLA+ specs for negotiation convergence

License

Copyright Taylor William Buley. All rights reserved.

MIT