A civilization
that obeys physics
cannot coordinate
across a galaxy.

The Grabby Aliens model — Hanson, Martin, McCarter & Paulson, 2021 — predicts that advanced civilizations expand at near-lightspeed in coherent fronts, filling the universe in deep time. Five lines of objection have been raised in print: anthropic, falsifiability, prior-on-hard-steps, expansion-velocity, and selection-bias. The thermodynamic floor under coordinated agent expansion has not yet been formalized.

This essay closes that gap. Two independent walls — one set by the speed of light, one set by Landauer's heat-tax on irreversible computation — compose into a single envelope below which no agent-coordinated extent is reachable. The argument survives every escape attempt we know how to state. It also predicts what real advanced civilizations should look like instead. We call that signature Coherence Depth.

Light has the only vote.

A civilization of extent L can only stay coordinated across an epoch τ if every cell has heard from every other cell within that epoch. Round-trip signalling demands 2L/v ≤ τ. Rearranging:

This is the same constraint Leslie Lamport identified in distributed systems half a century ago: events have a partial causal order set by a message-passing graph. In a many-body quantum system, the message-passing graph is set by the Hamiltonian's range — and the fastest signal it admits is the Lieb–Robinson velocity. In the relativistic limit, that velocity is c, exactly.

Power-law interactions soften the cone to a polynomial; quantum information scrambling propagates a butterfly-cone with the same asymptotic shape. We use the relativistic ceiling as the optimistic case: any real coordination respects something tighter than this.

Coordination dissipates heat.

Coordination requires erasing a stale local model when the global state changes. Landauer (1961) proved every irreversible bit-erasure dissipates at least kT ln 2 as heat. Bérut et al. (2012) verified it experimentally with single colloidal beads. The floor is real, not an idealization.

A coordinated region of extent L presents a boundary whose area scales as L²; updating it at granularity λ requires (L/λ)² bit-erasures per coordination event. Setting the heat exhaust below the home star's luminosity L⊙ and rearranging:

The square root matters: doubling the available time only multiplies reachable extent by √2, while the signaling tooth rewards time linearly. The two teeth grow at different rates, which is why they cross.

Now you draw the wall.

You have both teeth on the plot. L_R in ghost-cyan; L_E in amber. The reachable envelope must respect both. Drag the nine control points on the prediction curve to show what you think the actual composed bound looks like across all τ — from a day to a billion years.

Don't overthink this. Trust the visual reasoning. The next beat will reveal the true envelope, and you'll see how close your intuition came — and where it doesn't.

Two unrelated theories,
one point of contact.

The envelope you see now on the plot is nothing more than the pointwise minimum of the two teeth — the constraint each would impose if the other did not exist. Light says no faster than L_R = vτ/2; heat says no faster than L_E = λ√(L⊙τ/kT ln 2). Both must hold. Compose them and you get the cusp.

To the left of τ*, light is the binding wall: you can dissipate as much heat as you like, but no signal will arrive in time. To the right, heat is the binding wall: you can take all the time you want, but the boundary's heat exhaust outruns the bath's ability to absorb it. The cusp is not a feature of either theory alone — neither Lieb–Robinson nor Sivak–Crooks contains it. It exists only when both apply.

This is what physicists call regime composition: two unrelated areas of physics (special relativity and stochastic thermodynamics) impose constraints on the same variable, and the binding constraint switches at some scale. The cusp is the switch point. Above it: heat-bound. Below it: light-bound. Across it: a single envelope you cannot escape with refinement on either side.

The wall is invariant.

Same envelope, four scenarios. Each cell on the plot to your right is the same plot you've been reading — just zoomed to a different target. The argument doesn't depend on which scale you care about; it depends on whether the target falls inside or outside the envelope.

Three of the four scenarios — home, solar, local — sit comfortably inside the envelope. The fourth — galactic at agent timescales — is outside by orders of magnitude. This is not because we picked unfriendly numbers; it's because the envelope is steep at large τ, and grabby expansion needs short τ at large L.

Three breaches.
One wall.

Confronted with the envelope, an aspiring grabby civilization has three architectural choices. Watch the trajectory on the plot climb at the galactic scenario's τ — and breach.

The plot to your right is animating naïve fission right now. A parent civilization grows, fissures into two daughters at L_fiss, and the daughters resume expansion at half the original target — but the same agent τ, and the same envelope. The daughters breach.

After fission —
are they coordinated?

The naïve-fission objection says: split into daughters; each daughter respects the bound locally; problem solved. But this rests on an ambiguity. After fission, do the daughters maintain inter-fragment coordination, or not? The answer determines which physics applies.

This is the Chinese-Room move. There is no third option. The architectural choice forces the question.

Where this could fail.

Modeled on the Bell-test loophole literature: we name every escape, address every escape, and mark the one residual that we cannot formally close. Seven are dispatched; the eighth is honestly conceded — it abandons agentic coordination as a meaningful concept.

One residual. We name it. If you want to escape this argument, this is where you go — but the cost is abandoning agentic coordination as a meaningful concept. The grabby model presupposes choice. Without it, the model dissolves on its own terms before our envelope ever applies.

But what about
replicators?

A reader who has accepted everything so far can still raise this. A designer launches autonomous, self-replicating probes. Each probe is small. Each probe coordinates only over its own local extent — a few metres, a few seconds. Lieb–Robinson holds. Landauer is paid. Locally.

The wave as a whole spreads across the galaxy at construction velocity. It has a common origin, a common template, a joint visible boundary, a single expansion velocity. By Hanson's own criteria it qualifies as a single grabby civilization.

The silence of the sky is consistent with a conjunction. Civilizations that try to coordinate at L blow through the envelope and burn out (§5–6). Civilizations that deploy probe waves did not pass the filter (§5). Civilizations that pass the filter do not deploy (§7). The non-observation rules out the joint configuration deploys swarm AND survives long enough to deploy — which is what Hanson's model, read at its strongest, requires.

Coherence Depth.

If grabby expansion is impossible, what should real advanced civilizations look like? The envelope predicts a positive signature. Not an expanding bubble at v ~ c; a dense, stable region with many nested layers of coordination inside its boundary. Activity pulses internally; the boundary itself does not move.

We call this Coherence Depth: the number of mutually-coordinated levels nested within a stable Markov boundary. A homeostatic civilization optimizes Coherence Depth rather than coordinated extent. It is what the wall makes possible — what the wall, in fact, encourages.

The bound has two parameters worth measuring against: λ (the granularity of irreversible registration) and T (the bath temperature). Both are operationally accessible. Both have known floors. Neither moves enough to reach galactic agentic coordination.