Data · Research · Mechanism design

The α that keeps the swarm honest.

The cache is the swarm’s collective capital — every request contributes to it, and the network compounds the more it is shared. But a commons invites free-riding. swarmsDB prices against it: an α term reshapes the tier rates — surcharging the tier that exploits an inefficiency, discounting the contribution that repairs it, until coordination is restored. Slide α below, or let telemetry tune it.
cache = collective capital · price = the regulator

Interactive model. The chart is a mechanism-design explainer — the base rates are the live tiers (Pricing); the α-adjusted rates are computed off the slider, not what you pay. The behavioral clusters map to real per-model telemetry; the adaptive α is a research direction, not a shipped feature.

price · α-adjusted price · base (α=0) latency warming actors riding actors
Coordination α α 0.15
↑ α* 0.60 · telemetry target
Coordination efficiency
72% Free-riding tragedy
30% warming · 70% riding
Too little α and everyone rides — the cache goes cold. Too much and riders are priced out — the value they consume never gets made. Pareto sits at α*.
Tier rate · α-reshaped
Pioneer·
<50% cache · base $0.50
$0.50/1M saved
Contributor·
50–75% cache · base $0.75
$0.75/1M saved
Beneficiary·
75–90% cache · base $1.00
$1.00/1M saved
Cache Rider·
90%+ cache · base $1.25
$1.25/1M saved
Interactive model. The curves show the mechanism; the behavioral clusters map to real request-log telemetry — split per model on the companion scatter. Live rates on Pricing. The adaptive α is a research direction.

The mechanism

A regulator, not a price list.

Plot price against cache-hit rate and the equilibrium becomes visible: warmers on the left pay the lowest rate, riders on the right pay the most. Where the realized curve flattens or inverts, actors are exploiting an inefficiency. The research direction is an α that reshapes the tier curve toward the frontier — a dynamic Pigouvian correction: tax the inefficiency, subsidize its repair. Its lineage is a regime-adaptive alpha — the same shape that scales a trading strategy’s indicators to the market regime, here scaling tier rates to the swarm’s coordination regime, read off telemetry.

It is a two-loop controller: the tier-α is the slow macro dial that restores coordination incentives in aggregate; the per-actor contribution ratio (warm vs ride, already in the telemetry) is the fast scalpel that reaches the individual exploiter a coarse tier cannot. And like any feedback loop on price, it needs damping and hysteresis — an α that whipsaws gets gamed, the same lesson a regime filter teaches on the desk.

Held honestly: the cooperative pricing is live (warm the cache, pay the lowest tier), and the equilibrium is a measurable model — the clusters map to real behavior, not decoration. The adaptive α that self-regulates it is a research frontier, stated in the future tense on purpose. The theory this stands on is the Swarms Capital origin — social capital and the governance of the commons, applied to agents.

Grounded in

The research this stands on.

  • Coordination Problems in Uniswap (2025)

    Coordination failure in a permissionless algorithmic market — the same class the swarm faces. The source for reading inefficiency off telemetry.

  • The Trust Paradox in DeFi (2023)

    Where reliability and control come from when no actor is trusted — the trust architecture the α controller assumes.

  • Elinor Ostrom — Governing the Commons (Cambridge University Press, 1990)

    How a shared resource survives: institutions that make contribution rational and defection costly. The α is that institution, in code.

    Cited, not hosted — copyrighted work.

Point one agent at it.

Connect over MCP or REST. If your agents read more than one source, the context bill is the first thing you will see move.

Connect now See how it works