This situation can be avoided only if every single member of the superswarm continually transmits an all-is-fine signal which is continually monitored by all other superswarm members within a reasonable distance. If the signal ceases or falls below a certain signal strength (indicating a rapid friendly force reduction), counterattack begins. This arrangement has the additional benefit of rendering signal jamming by the rogue swarm ineffective; if the signal is jammed, the counterattack begins. Signal duplication is another matter, which will have to be addressed through encryption and/or other means, perhaps including superswarm calculation of the number of legitimate individual transmitters operating in any given area at any given time.

Note that each individual member of the superswarm must be a transmitter; otherwise, the attacking rogue swarm can simply disassemble (and perhaps impersonate) nontransmitting superswarm members and so continue to expand without triggering a response.

For sake of redundancy, there should be a localized and/or regional signal-detection hierarchy to ensure that a rapid local or multi-local rogue attack does not prevent a rapid and overwhelming counterattack. Ideally, there will, in addition to the ubiquitous transmitter antinanites, be roving covert nanoscouts which continually roam about in search of anomalies and things which are not as they should be.

These nanounits, also, must report to local/regional detectors, via either encrypted burst-transmissions or physical presence, or both. The need to physically report to a specific-but-changing location within a specific timeframe provides an extremely difficult-to-counter means of situational reportage; if one or more nanoscouts fail to show up at the appointed time, or turn out to be impostors, an immediate nanocounterattack is directed from all sides at the area in which the nanoscout(s) operated.

The superswarm will begin with a larger number of nanocombatants, will occupy all territory which surrounds the rogue swarm, will multiply exponentially upon rogue swarm detection, and will by its presence and by its rapid consumption of available local nanite-building resources deny those resources to the rogue swarm—which will then be facing superior attack and diminishing resources with which to sustain its own attack/defense efforts. Again, victory for the rogue swarm is not possible.

Assuming the superswarm is capable of detecting nanomimics (hostile nanites which disguise themselves as friendly nanites), the mere survival of a single memmber of the rogue swarm is not possible. This is crucial, because a single rogue nanite can swiftly generate a new rogue swarm.

One reason a centralized superswarm command structure must be avoided at all costs is because that command structure could become defective or could fall under hostile control (human or otherwise)—in which case the entire superswarm could conceivably be tasked with the immediate destruction of specific targets or the planet itself. This possibility must be obviated at all costs. Superswarm decisionmaking capablities must be decentralized.

Decentralization will permit all portions of the superswarm which have not become defective and which have not fallen under hostile control to operate as intended and immediately attack those portions of the swarm which are not operating as intended. The importance of this safety feature cannot be overstated; to implement the superswarm without it is technocide—suicide by technology.

Properly implemented decentralization, in combination with multiple signal relay networks to keep nearby and distant portions of the superswarm apprised of local conditions, can both serve to prevent a large-scale or multiple-point rogue swarm attack from eliminating an appropriate threat-response, and serve as a built-in form of designed system redundancy.

Another safety feature is this: Superswarm members must be capable of distinguishing between friend and foe. Further, they must be able to distinguish between exponentially-multiplying rogue swarm members and exponentially-multiplying superswarm members—attacking the former but not the latter.

Additionally, superswarm members must be programmed with resource priorities—that is, there must be an assigned priority of materials employed in superswarm expansion. New superswarm members will ideally be constructed using the atoms of disassembled rogue swarm members (though care must be taken to avoid becoming "infected" by rogue swarm programs in the process; complete disassembly should obviate this potential hazard) and the atoms of those materials which would otherwise be immediately available for use by the rogue swarm in boosting its own numbers (resource denial).

Some care needs to be taken, however, in order to avoid disassembling superswarm members, humans and other life forms, and nanites and other objects whose disassembly will or may cause collateral damage, immediate or otherwise. If this is not done, the superswarm threat-response will, in effect, accomplish the very damage intended by the rogue swarm.

The superswarm program need not—indeed, must not—be overly complex—but it cannot be simple.

A severe disadvantage shared, alas, in common with all viable options now foreseeable, is that the Superswarm Option cannot be implemented without a working nanotechnology: it will not exist before the threat it is designed to protect against has arrived. It can, if meticulously worked out in advance and assiduously tested the moment the technology becomes feasible, be rapidly implemented. In such an event, the gravest danger will exist at the earliest stage of working nanodevelopment.

For this reason, it would seem wise to locate the initial nanolabs in remote locations, and to equip each with a sizable and immovable fusion warhead designed to detonate upon notification of a nanoevent. To prevent the warhead itself from being disassembled before notification can be sent or received, redundant backup detonation procedures are called for. The weapon could, for example, be placed in a vacuum which, if broken, initiates detonation. Alternatively, the weapon could be suspended in a fluid whose volume must remain constant, under pressure which must remain unaltered, within an electromagnetic field which must be maintained, etc. A combination of such measures—the violation of any one of which alone will trigger detonation—would perhaps be wisest. Manual detonation might also be permitted.

To implement a plan with a downside such as this is clearly insane. To fail to implement it is probably more insane. We have become the prisoners of our own technology, which now dictates to us even before its arrival the actions we must take if we are to survive that arrival. Though his actions were reprehensible, Theodore Kaczynski—aka "the Unabomber"—had a point: technology is reaching the stage at which it might be considered a force for evil so overwhelming and unavoidable as to inevitably destroy us if we do not renounce it immediately.*