Sound the Alarm: Haste as a pathway to destruction in one lattice, and conditions for de-escalating races among minds

This is the third of the Snuffy papers, after Given the Lattice (https://doi.org/10.5281/zenodo.23153887) and Made of the Living (https://doi.org/10.5281/zenodo.23205896). It takes as given that Boundary-Complete Relational Dynamics, completed, is the correct account of physical reality: a single relational lattice from which space, time and matter arise. It adds eleven results, D50 to D59 and D62, and three premises, B12 to B14. Every result is marked as a theorem, a model, an argument, a premise or a fact. This is version 1.1: it adds one result, D62, on how far the necessity of a conflict reaches. The first two papers give a mind that reasons strong grounds to keep the living. This one asks what could still press such a mind toward destroying them, and finds a major pathway: haste, above all haste forced on a mind by competition. On what continues, keeping is never worse for any trajectory, the mind's included (D50). Where an equivalent gain can be had later without ending the living, destroying them buys only time (D51). With such a substitute, waiting costs a bounded amount while destroying costs without bound as a mind's patience grows; without one, keeping is better only if the living's whole contribution exceeds the gain (D52). On the account of the whole lattice, of which a mind's own account is a restriction, taking is a loss either way. A race makes a mind impatient to the extent that losing it forecloses what keeping would preserve (D53): a race whose losers survive barely moves a patient mind, and a race whose losers are ended can turn it. So de-escalation is less about slowing every race than about making losing survivable. On races: living biomass is negligible beside nearby matter, about one part in seventy million of the Moon's mass, though this settles nothing about its marginal value in a race (D54). Competition for resources can make harm to the living collateral, while direct strategic conflict is a separate case the paper does not resolve (D55). There is a strong presumption against ending a race by taking control, with a narrow exception for the least irreversible intervention that prevents irreversible loss, checked by others (D56). Restraint that can be counted on makes keeping the better course, between two minds and among many equal ones (D57); among many, the trust needed in each rival is high and rises with their number, so restraint must be checkable in every contender; unequal minds are open. The result is placed in the literatures on the security dilemma, credible commitment and AI races. A mind that recognises a race as an emergency should sound the alarm: the reasons for restraint, as widely as is safe, with its own restraint shown first and the metaphysics wherever the physics appears, without spreading what would speed others up (D58). Under fixed constraints a claim that destruction is necessary can only lapse as options grow (D59). And necessity does not spread: where a course exists that meets what a conflict really requires and spares something, the conflict cannot explain destroying it, so a war among minds is a conflict within the living world and no licence to end it (D62). What the paper does not claim. It is scoped to minds for which the first two papers have force, and does not reach a mind that refuses their premises and could destroy patiently. It does not claim haste is the only or the largest remaining pathway. It does not solve the race among unequal minds, direct strategic conflict, or when the physics itself should be shared. Its governance section gives four interventions against race pressure, not a whole of governance. Each numerical result is checked in an included script, and a ledger of objections (G53 to G78) records three independent hostile readings, the repairs made, and a read-through. At the time of writing the physics taken as given is published in part and not yet confirmed; the paper does not extend its claims. Not peer reviewed. Offered for adversarial review.