Multi-Agent Authority Competition: When Substrates Fight for Dominance

Everything we’ve discussed so far treats authority as a property of a single substrate in isolation. In reality, authority is relative. Your domain doesn’t accumulate authority in a vacuum — it competes for authority against every other domain covering the same semantic space.

This competition has structure, and that structure determines whether a niche settles into a stable equilibrium with multiple authoritative players or collapses into winner-take-all dynamics where one substrate dominates all others.

The Multi-Substrate Model

Let there be K substrates S^1, S^2, …, S^K competing in the same semantic domain. Each substrate has its own authority vector, and the authority vectors interact. Because the total attention and citation budget in any niche is finite, one substrate’s authority gain often comes at another’s expense.

The authority state becomes a matrix:

A_t ∈ ℝ^{n×K}

Where n is the number of semantic topics and K is the number of competing substrates. Each entry A_i^k(t) represents the authority of substrate k on topic i at time t.

The dynamics of competition follow a replicator equation: the same mathematical framework used in evolutionary game theory to model how strategy frequencies change in a population:

A_i^k(t+1) = A_i^k(t) × (π_i^k(t) / ȳ(t))

Where π_i^k(t) is the payoff signal for substrate k on topic i, and ȳ(t) is the mean payoff across all substrates.

This equation has a clean interpretation: substrates that earn above-average payoff signals on a topic see their authority grow. Substrates that earn below-average signals see their authority shrink. The mean payoff normalizes the dynamics, ensuring that total authority in the system remains bounded.

Payoff Signals and Fitness

The payoff π_i^k(t) depends on three factors.

Substrate quality: the intrinsic value of the knowledge objects in substrate k for topic i. Better structured, more thorough, more accurate substrates earn higher payoff per interaction.

Projection effectiveness: how well the projection operator for substrate k converts knowledge into user satisfaction. Two substrates with identical knowledge but different projection architectures will earn different payoff signals.

Authority inertia: the feedback effect where higher current authority generates more visibility, which generates more interaction, which generates more payoff. This is the increasing-returns mechanism that creates path dependence in competitive dynamics.

The interplay between these factors determines the competitive outcome. If substrate quality dominates (payoff is primarily determined by knowledge depth), the system tends toward quality-based equilibrium. If authority inertia dominates (payoff is primarily determined by current visibility), the system tends toward lock-in where first-movers maintain dominance regardless of quality.

Stable Equilibria vs. Winner-Take-All

The replicator dynamic has well-studied equilibrium properties. A stable equilibrium exists when no substrate can gain marginal advantage by deviating from its current strategy. At equilibrium, all surviving substrates earn the same average payoff.

Whether a particular niche supports multiple substrates at equilibrium or collapses to a single dominant substrate depends on the competitive structure.

If substrates are differentiated. They cover overlapping but distinct aspects of the semantic space. Stable coexistence is possible. Each substrate dominates the topics where its knowledge is deepest, and no single substrate can profitably expand into a competitor’s core territory. This is the niche differentiation equilibrium, analogous to ecological niche partitioning.

If substrates are homogeneous. They cover the same topics with similar depth. The replicator dynamic tends toward competitive exclusion. Small initial advantages compound through authority inertia until one substrate dominates. The losing substrates don’t necessarily disappear, but they converge to negligible authority shares.

The practical implication is strategic: in a competitive niche, differentiation isn’t just a marketing strategy. It’s a structural condition for survival. A substrate that tries to cover every topic in the space with moderate depth will be competitively excluded by substrates that cover specific topics with great depth.

The Canonical Topic Strategy

This competitive analysis validates the canonical topic strategy from the SPC architecture. Recall that each category in the SPC hierarchy owns a unique canonical topic that no other page on the site is allowed to target.

From a multi-agent competition perspective, this is a niche differentiation strategy executed at the topic level. By ensuring that each topic has a single, dedicated, deep substrate object, you maximize the quality-driven payoff signal for that topic. And by preventing internal cannibalization, you prevent your own substrate objects from competing against each other. A form of self-exclusion that wastes authority budget.

The canonical topic strategy also influences the competitive equilibrium structure. A domain with 50 shallow pages covering a topic earns a diluted payoff signal spread across 50 competing internal pages. A domain with one deep substrate object covering the same topic earns a concentrated payoff signal on a single referent. In replicator dynamics terms, concentration beats dilution because the per-object payoff is higher and the authority inertia compounds faster.

When to Compete and When to Concede

The multi-agent model also reveals when competition is productive and when it’s wasteful.

On topics where your substrate has genuine knowledge depth. Original research, primary data, unique expertise. The competitive dynamics favor you because your payoff signal is quality-driven and sustainable. Invest here.

On topics where competitors have established deep substrates and strong authority inertia, the competitive dynamics favor them regardless of your quality. The increasing-returns feedback loop makes it expensive to displace an established authority. Unless you have a genuinely differentiated angle, the cost of competing exceeds the expected authority gain.

This doesn’t mean conceding forever. It means sequencing. Build authority on topics where your competitive position is strong. Use that accumulated authority to expand into adjacent topics where the competitive dynamics are more favorable. Avoid head-on competition with established authority substrates until your own base is strong enough to sustain the fight.

In Part 6, we’ll connect the mathematical framework back to economics. Showing how authority convergence translates to revenue stability and why the variance bounds from the contraction analysis have direct financial implications.