Proof of Indexing: Agent-Based Simulation

Simulation code for the paper "Proof of Indexing: Indexer-Level Staking for Shared Route Graphs" (Tham, 2026).

Validates the passive slashing mechanism described in the paper by modelling indexers, routes, automatic execution feedback, and stake-based accountability over 24 months.

Quick Start

pip install numpy
python run.py

What It Tests

1. Healthy equilibrium — With all rational indexers, zero slashing occurs. The threat is sufficient.
2. Negligent indexers — Bad actors are slashed and demoted. Good indexers are unaffected.
3. Competitive sabotage — Attacking a competitor's stake costs real money and gains nothing (burn, not reward).
4. Parameter sensitivity — Sweep across stake sizes and slash fractions to find viable operating regimes.

Parameters

From the companion paper "Internal APIs Are All You Need" (Tham & Garcia, 2026):

Parameter	Value	Source
Fee per lookup	$0.01	Main paper ($0.005–$0.02 range)
Contributor share	70%	Main paper
Graph size	500+ domains, ~10K endpoints	Main paper
Verification cadence	Every 6 hours	Main paper
Lookups per route per day	5	Estimate
Monthly failure rate	5%	Estimate
Maintenance cost per route	$0.50	Estimate
Feedback threshold (k)	3 reports	Design parameter
Grace period	1 month	Design parameter

Files

• poi_simulation.py — Core simulation engine (Route, Indexer, Graph, Simulation classes)
• scenarios.py — Pre-defined test scenarios
• run.py — CLI entry point, runs all scenarios and prints results

Key Finding

In the healthy equilibrium, no FDRY is ever burned. The threat of slashing is sufficient to produce rational maintenance. This is how performance bonds work in the real world.

Citation

@techreport{tham2026poi,
  author = {Lewis Tham},
  title = {Proof of Indexing: Indexer-Level Staking for Shared Route Graphs},
  institution = {Unbrowse AI},
  year = {2026}
}