Space Warfare Seems Mostly Defense Dominant

Epistemic status: I’ve done some thinking and research on this but I am not a physicist and I can easily be wrong about specific things.

Sometimes when thinking about the long term future, it is interesting to think about the offense-defense dynamics of a fully colonized ‘mature’ universe. E.g. suppose we succeed at aligning AIs and colonize a large section of the universe but there are some unfriendly aliens who have colonized the rest, or alternatively that we have colonized our galaxy with aligned AIs but there’s a bunch of rogue AIs running around in the other galaxies, how worried should we be? More generally, in the longer term evolution of the universe there is going to be a long period after colonization, but before all other galaxies become unreachable due to cosmic expansion, that there is wide room for ‘interactions’ across cosmic distances. Will the universe end up mostly or entirely peaceful with everybody just quietly maximizing utility as seems best to them or will there be e.g. more classic SciFi style interstellar or intergalactic conquests with e.g. ravenous hordes of replicators eating galaxy after galaxy?

To think about this, we’re going to try to assess what a hypothetical intergalactic war between two roughly equivalent galaxies would look like, where both host a mature AI Kardashev-3 style civilization – i.e. one that has maxxed out its ‘tech tree’ and can harness the mass/energy of its entire galaxy largely as it pleases. We are also going to assume no dramatic departures from known physical limits – i.e. no wormholes, time-travel, infinite-energy-glitches or exceeding the speed of light, since if any of these are possible then all bets are off¹.

My general conclusion after thinking about this a fair bit is that in the end the dynamics are fairly strongly defense dominant, if both adversaries are equally capable and paranoid, in that an attacker would need to spend far more mass/energy to destroy the defender than a defender would need to expend to defend itself. An attacker is extremely unlikely to ever be able to fully wipe out a mature and prepared defending K3 civilization, although they could potentially seriously cripple them if the defender is taken by surprise or not careful, and that in general such intergalactic wars would last for millions and potentially billions of years, would be ruinously expensive and unpleasant for both sides, and would permanently hurt the long-term growth potential of both attacker and defender.

This makes large scale war between mature K3 civilizations ‘irrational’ in the sense that the attacker would likely spend more resources in the war than they could expect to gain by e.g. colonizing the remnant of the defender’s galaxy after they have been wiped out. However, there are exceptions to this which could still drive war. Firstly, if the attacker has an ‘irrational’ dislike of the defender civilization for e.g. ideological or other reasons this could motivate a prolonged campaign which does not pass purely economic cost/benefit considerations. Secondly, if the attacker assesses the defender to be a young or ‘inexperienced’ budding K3 civilization then they could be highly vulnerable to a first strike and obliterated much more cheaply than a mature and highly paranoid K3 would be. Intergalactic wars (unsurprisingly) seem likely to be highly visible to the rest of the universe which could prompt interventions, although over timescales again of millions of years.

To see how we get to this conclusion, let’s wargame out what this theoretical war would look like. For illustrative purposes let’s imagine that we are a Kardashev-3 civilization inside the Milky way and we are defending against a similar civilization in Andromeda, roughly 2.5 million light years away. Let’s assume both galaxies are roughly ‘evenly matched’ (i.e. to within an OOM) of energy density, number of stars, mass etc. Let’s break down what the offensive and defensive capabilities are for a sensible K3 civilization.

Offensive weaponry:

These come in three main classes:

First, pure electromagnetic beam weapons. Specifically, these will be laser or maser beams of incredibly high power and extremely well collimated which can cross intergalactic space to hit specific spots at the target galaxy.

At K3 scale, we can get extremely powerful with these. Around each star we assume there will be a Dyson sphere and stellar infrastructure capable of constructing massive phased arrays which can operate in perfect synchrony. With such arrays, it is well within the known laws of physics to create highly collimated lasers which can converge on relatively small targets (e.g. perhaps a few hundred kilometers) across vast intergalactic distances. This process is highly energetically efficient since the intergalactic medium is almost completely empty, and hence transparent across almost all frequencies. Surprisingly, because of this, once the laser is fired, the actual transmission itself will be almost lossless and can roughly transfer 1 watt of power at source to ~0.999 watts of power at the destination. Naturally you might think there would be large losses due to scattering and poor collimation of the beam, but with a wide enough aperture at the source these problems can be driven down to extremely manageable levels. Certainly building such apertures thousands of kilometers across pose engineering challenges but seem extremely straightforward for our hypothetical K3 civ. The only real energy inefficiency comes from the power management and focusing of the laser at the source.

The infrastructure for building an intergalactic laser around a Dyson sphere is not dramatically different from ‘civilian’ infrastructure needed for other purposes such as e.g. peacetime intra-galactic power transfers, and so it is not an obvious sign of hostile intent either but can just be regular civilian infrastructure until repurposed.

The ultimate strength of such lasers can be extremely powerful. Roughly each Dyson sphere could collimate a beam of a significant fraction of their stars energy if desired. Moreover, with sufficient advance notice, these beams could be coordinated galaxy-wide to hit the same point in the defender’s galaxy in a relatively straightforward way (for a K3 civ). This means it is possible to concentrate galactic level energy ~10^35 watts at relatively small regions of e.g. a few hundred kilometers radius anywhere within the defender galaxy as desired. This level of energy would be expected to almost instantly destroy any target it hits. To give an idea of how powerful this is, the gravitational binding energy of the earth is approximately 10^32J. Our hypothetical intergalactic laser, if focused on the earth, could utterly vaporize it in only 10 microseconds. These lasers are likely highly reconfigurable and could theoretically hit tens to thousands of separate targets every second – note that at a distance of 2.5M light years, the change in angle needed to hit different targets within the same galaxy is miniscule.

There is also no warning from the intergalactic laser. The photons travel at the speed of light. The first photon you get warning you of the attack is the attack. This is unlike other methods involving kinetic transport of mass between galaxies which necessarily travel less than c and hence give warning time.

If this laser is so powerful and there is no warning, what are the possible drawbacks and countermeasures? The principal issue is simply time caused by the immense distances involved. If the laser hits, it is devastating, but can it hit? When targeting the laser you are operating from targeting information 2x the distance in light years old. E.g. to hit the Milky Way from Andromeda your laser would be trying to hit targets which were last seen 5 million years ago. 2.5 million years for the photons to travel from the Milky Way to Andromeda, so you can see them, and then another 2.5 million years for your laser to head back.

A lot can change in five millions years, especially against a paranoid adversary. The orbits of stars, planets, and the supermassive black hole at the center of the galaxy, as well as broad galactic features are highly predictable at this timescale, especially with the level of in-depth astronomy and interferometry a K3 civilization is capable of. However, more mobile targets such as habitats, and perhaps planets, are generally much harder to hit as even small uncertainties in delta-v compound into massive potential volumes after 5M years. Similarly, if the adversary is expecting an attack and randomly moves their infrastructure, you very quickly run into a vast potential volume to cover and although the laser is extremely powerful when it hits, trying to blanket an enormous volume of space stochastically will explode your energy budget. Space is very very big and almost all empty. Conversely, you can spread out the volume hit by the laser but this proportionally dilutes the power until it is no longer a death ray but more just a moderate warming applied over a huge area.

So, the intergalactic laser is exceptionally powerful and efficient against fixed infrastructure on predictable orbits. If you are targeting a naive K3 then this will include almost all of their Dyson-sphere based infrastructure, since the orbits of stars are extremely predictable and stars are expensive to move. It would also include any built up infrastructure in the galactic core around the SMBH, which is clearly not movable, and potentially any other interesting galactic features².

However, what if you are a paranoid K3 defender? In this case you do not put the majority of your infrastructure in static extremely predictable orbits. You pay some price of inefficiency to do so but this allows you to be much safer from this kind of intergalactic sniping. What infrastructure you do have around stars is treated as sacrificial rather than core. You pursue starlifting to spread out and uniformly distribute the hydrogen rather than keeping it burning in stars. Billions upon billions of habitats flying around the galaxy on irregularly and randomly shifting orbits each with many billions of years of backup hydrogen to fuse is much much safer in war than a bunch of fixed heavily concentrated Dyson-sphere based infrastructure around bright and mostly immovable stars.

In a pinch, you can also use black holes as shields from intergalactic lasers as well. If you know the angle you want to protect against (e.g. you are only defending against one specific adversary rather than potentially the entire universe), you can situate core infrastructure in the ‘shadow’ of the black hole along that line of sight. The intergalactic laser aimed at this region will then either directly enter the black hole, causing it to expand marginally but with no other ill effects, or will lens around it and pass onwards, missing the targets in the black hole’s shadow. The size of such regions is not massive though and being in such close proximity to a SMBH likely has other issues.

In sum, intergalactic lasers are both relatively trivial to build (for a K3) civilization and also capable of immense and energetically efficient devastation of the opposing galaxy. Such beams cannot be seen before they hit and cannot reasonably be stopped³. The fundamental limitation of such lasers is simply the vast and inescapable targeting delays imposed by the speed of light. While being capable of absolutely devastating unprepared K3s which cluster around fixed and predictable points such as stars, against mature and paranoid K3s which exist much more diffusely throughout their galactic region, being comprised of uncountable ‘subunits’ each on their own randomly shifting orbits, then the power of lasers is very much diminished. Certainly, even by stochastically raking through space inside the ‘high density’ parts of the opposing galaxy, they will score some hits, but the power efficiency of such a scheme would be dramatically diminished.

The second weapon type is relativistic kill vehicles (RKVs). The general idea here is to shoot objects at relativistic speeds e.g. 0.9 or 0.99c towards the enemy galaxy. The kinetic energy at such speeds is enough that even without explicit warheads, the simple collision unleashes immense energy. RKVs can come in a huge range of sizes from large e.g. asteroid or moon sized objects to tiny golf-ball or dust sized. A generally sensible design would be for medium sized RKVs to be sent out from the source galaxy which then would break apart into many separate small warheads close to the destination to saturate the space close to the target.

Compared to beam weapons, RKVs are much less efficient energetically. Substantial energy must be spent accelerating the RKVs up to relativistic velocities (although we would presumably do this by pushing with lasers, magnetic sails, or some kind of railgun so we do not have to pay the rocket equation tax.). Moreover, the RKV itself is also mass which is lost to the source galaxy forever. Unlike beam weapons, RKVs also necessarily travel slower than light, which means that the defender can have a have a reaction time depending on how early they detect the RKV (either in launch or in flight) and the speed at which the RKV is travelling.

RKVs do possess one massive advantage over beam weapons though, namely that since they are physically closing the distance, they can retarget based on up-to-date information instead of using positions that are millions of years out of date. Instead the RKV can utilize on-board sensors to get increasingly up to date information as they approach and ultimately enter the target galaxy. When launched they are initially sent at some broad region of the galaxy but then as they close in, perhaps within the final ten to one hundred thousand years of flight time, they can perform up to date surveys and mapping of the infrastructure inside the defender’s galaxy and choose targets appropriately. If equipped with onboard thrusters, it does not take much lateral delta-v to narrow in on a trajectory targeting some specific orbit or habitat. Ultimately, with a basic navigation and closed loop control software, which would be trivial for a K3 and could even be made today, an RKV could continuously home in on a preselected target right up until the moment of impact. If you want to hit mobile targets such as habitats in practice, RKVs are the weapon you need.

What is more, the energy to launch RKVs is trivial individually for a K3 civilization and indeed using a moderate fraction of a galaxy’s energy budget it is possible to match trillions or quintillions of RKVs simultaneously at a target. Likely, such a bombardment would be almost continuous from the attacker to defender galaxy over over millions of years and cover a vast range of RKV sizes and types from extremely large asteroid sizes RKVs to take out massive individual pieces of infrastructure such as planets or large-scale megastructures, to much smaller ones to target individual freely moving habitats.

The downside of RKVs, however, is that, while they are relativistic, they still move slower than light. This means that the defenders get reaction time which they can use to deflect, disable, or destroy the RKV before it hits its target. Supposing an RKV is launched at 0.9c, the maximum reaction time is 250000 years if detected at launch, which is clearly extremely long. If the RKVs are detected during intergalactic transit this could be tens or nearly a hundred thousand years. Even if the RKVs are close and already inside the defender’s galactic halo, reaction times of decades or centuries are still possible. Even if not detected until somehow inside the solar system of the target, there is still a reaction time of minutes to hours if detected on the outskirts.

This then brings us to two questions. How detectable are RKVs and then, once detected, how easy is it to thwart them.

Detectability is an adversarial game. Attackers can make RKVs naturally hard to detect by keeping them extremely cold and dark, by radiating away all heat in a collimated beam away from the targeted galaxy, by using regular stealth techniques to try to fool radar and other systems, and by only using passive sensing until close to the point of impact.

To counter this, defenders can create a galaxy wide sensor mesh, utilize incredibly large and powerful telescopes (literally built on a galactic scale) to scour the surrounding space for any abnormalities, and build networks of radar and other sensor outposts in spheres around their core galactic region potentially extending hundreds of thousands to millions of light years. Much of this infrastructure even peaceful K3 civilizations will want to do to protect against other miscellaneous natural threats and also to answer scientific questions.

RKV launches are likely not that energetic compared to all the other things a K3 civilization is doing and thus can likely be somewhat hidden in the general ‘noise’ of K3 industry of the attacker galaxy. However, in intergalactic space they are fundamentally unusual objects – being both substantially bigger than the ‘regular’ intergalactic dust and also travelling at vastly greater speeds than almost all natural objects. This makes this an equilibrium that seems likely to tend towards the defender vs the attacker, although this is not guaranteed. Once an RKV hits the galactic halo it will start encountering much higher dust and other matter density than in the intergalactic void which will inevitably cause some discharges and disruptions which an alert and paranoid K3 civilization could certainly notice. Paranoid defenders could also deliberately create diffuse dust ‘minefields’ along likely attack vectors to their galaxy which would both slow and potentially damage incoming RKVs as well as light up their positions for other defenses to see.

Now, assuming an RKV can be detected, how can it be stopped? The crucial insight is that the defender does not need to vaporize or otherwise destroy the mass of the RKV and dissipate its kinetic energy (which would be large and expensive). They only need to nudge it slightly off course. Because space is almost all empty, once off-course the chance of an RKV hitting anything is very small. This means that the defender only has to kill or disable the RKVs sensor and movement capabilities as well as change its course by some equivalently small delta-v as the RKV itself has to use to home in on a particular target.

Dyson-sphere-based or other large beam weapons are perfect for this. Large scale laser infrastructure is extremely useful for defense as well as intergalactic death rays. Moreover, the lasers do not need to completely vaporize the RKV but simply ablate the outer surface, blind the sensors (which are likely on the outer surface for obvious reasons) and provide small but significant delta-v nudges to push the RKV onto a harmless path. A dyson sphere based laser defense setup, when combined with detailed tracking and targeting data from large-scale telescopes inside the defender galaxy as well as sentinel outposts extending out far into the galactic halo and beyond, could plausibly track and engage hundreds or thousands of incoming RKVs per second with a massive energy disparity between the energy needed to shoot the RKV in the first place and the energy needed to deflect one on the defense. The RKV defense setup also removes the key disadvantage of intergalactic lasers – that of targeting delay. During RKV defense, as the RKV approaches your targeting delay gets smaller and smaller. The RKVs can likely make random small trajectory adjustments to create a probabilistic volume that must be filled with laser shots to guarantee a kill, however at a small enough distance away this volume can get quite small and the RKV itself, limited by the rocket equation, has finite delta-v it can expend on random jitter. The static dyson-sphere based laser infrastructure can keep firing at a high clip forever. In the equilibrium, all things being equal, the defense seems favoured here.

The third type of weapons are invasion ships – this is the classic science fiction trope, however actual invasion ships have one fundamental weakness – they need to slow down at the destination galaxy. This has two effects. Firstly, energetically getting invasion ships to the opponents galaxy is substantially less efficient than sending RKVs there. This is because of the tyranny of the rocket equation. While the invasion ships can be accelerated to relativistic velocities at origin galaxy, to slow down, it cannot be assumed there is an equivalent infrastructure at the destination. Instead, the invasion ships must carry their own braking fuel with them, which must then also be accelerated and so on.

The second fundamental problem is lack of stealth. When accelerating your exhaust points away from your target, when decellerating your exhaust points towards it. Essentially your are deliberately dissipating all your kinetic energy as a gigantic beacon screaming ‘I am here come kill me’. The decelleration burns of large-scale invasion fleet would both likely last thousands of years and also be immensely noticeable to any reasonable civilization in the target galaxy let alone a paranoid K3. Even if you don’t try to decellerate by rockets but instead by e.g. drag on magnetic sails, this drag causes friction which then radiates uniformly in all directions, again serving as a beacon.

A more minor problem is that the invasion ships themselves are also likely significantly more delicate than RKVs since RKVs simply have to hit a target and are basically pure slugs used solely for their kinetic energy. The invasion ship by contrast is likely highly engineered and needs the capability to autonomously stand up a military self-replicating civilization capable of fighting off whatever elements of the defender civilization exist wherever they land. This means that potentially even ‘minor’ damage that an RKV could shrug off could cripple an invasion ship and also that each invasion ship is likely nontrivially more expensive than the RKV to construct.

This is inevitably going to be an extremely tough battle however. Unless almost entirely obliterated beforehand, the defender civilization is likely heavily entrenched with all its productive infrastructure right there and ready to go. The invader needs to somehow either carry that with them (extremely expensive due to rocket equation) or somehow construct it once there (difficult because the defender has presumably had millions of years to gather and colonize all easily accessible resources). Similarly, the attacker has supply lines that are millions of years long (!!) while the defender’s supply lines are only perhaps decades or centuries for interstellar transport. Moreover, the defender has had potentially thousands of years to prepare during the decelleration phase, even if for whatever reason they just waited for the invasion fleet to arrive vs attacking it during the decelleration burn.

Sensing capabilities

It is also worth briefly digressing on the potential sensing capabilities of a mature K3 civilization. Even on Earth today we are capable of performing nontrivial mapping of the universe and detection of distant galaxies. Any mature K3 would have interferometers with apertures effectively spanning multiple AU if not light years, allowing them to construct detailed maps of every star in the universe, peer back into the most distant reaches of time, and perform extremely detailed mapping of any systems of interest especially in the ‘near’ universe. To put some concrete numbers on it, an interferometer that could resolve a 1km region in Andromeda (2.5 million light years) would be about 100AU in diameter (approximately 3x the orbit of Pluto). This would not be a single dish but constellation of regularly spaced and interlinked satellites with smaller telescopes, which synchronize their data to build a unified picture. For a K3 civilization, this kind of infrastructure is trivial and the K3 would likely have millions if not billions of these constructed throughout its galaxy (concentrated around the edges). With this kind of resolution, detailed mapping of many important star systems and other infrastructure would almost certainly be done on any neighbouring K3 civilizations.

As briefly discussed in the RKV section, a mature and paranoid K3 civilization will also invest very heavily in maintaining extremely high levels of situational awareness within its own galaxy and especially around the edges and in ‘near’ extra-galactic space. While physically tracking literally every single hydrogen atom coming in and out of the galaxy is likely unfeasible, something tracking any objects of greater than some relatively small size, as well as those with unusual spectra, velocity, trajectory, or other suspicious feature is certainly doable given sufficient, and relatively small, effort. This global tracking would likely also be complemented with stochastic sampling of much smaller objects. In extra-galactic space, the problem is even easier since matter is so diffuse. Almost all objects of nontrivial size could be tracked using radar and other methods out a reasonable distance from the galaxy. A diffuse web of deep-space probes could be trivially constructed that extend far into extra galactic space which can perform detailed monitoring of their own environment, powered by beamed power from the host galaxy with minimal efficiency loss.

K3 attack and defense doctrines. Putting this all together, we can try to get a sense of what offensive and defensive doctrines look like for our mature K3 civilizations. On the offense, the plan would look something like:

0.) Monitoring and mapping. In the millions of years prior to the conflict, the aggressor K3 would build up exceptionally detailed map of the defenders civilization including all population and industrial hubs and other critical infrastructure as well as carefully monitor defensive installations and e.g. predictability of their movements.

1.) Simultaneous first strike using primarily beam weapons on core infrastructure and all targets that have been identified during monitoring. This will include all infrastructure around the SMBH, all dyson spheres around stars, all planets which have not been diassembled and which appear to have heavy infrastructure, as well as stochastically raking through regions of space which appear to have high concentrations of industry or energy harvesting or use.

2.) Simultaneously with the first strike we would want to hit with a vast wave of RKVs to target and destroy the core infrastructure which is too mobile to hit with beams directly. This would likely include concentrations at core industrial regions as well as just generally saturating all populated areas from as many angles as possible. These would be timed carefully so that they hit in synchronized waves with the goal of overwhelming defensive laser installations. The RKVs would be launched thousands or millions of years in advance of the laser strikes to precisely hit at the same time. They would employ advanced stealth for their intergalactic voyage as well as e.g. electronic warfare countermeasures, decoys, chaff, etc to try to fight back against the defensive stations trying to intercept them. Once they near their destination they would split apart into a volley of smaller projectiles weaving randomly which can saturate the volume of space near their targets. These smaller projectiles would be harder to track and hit with defensive lasers and would have engines capable of imparting substantial delta-v, requiring defenders to saturate large volumes with laser first to guarantee a kill at moderate distances – i.e. light hours or more.

Initial targets would then be core industrial, defensive, and sensing infrastructure. I.e. if you can eliminate the ability of the defender to observe track future and deflect RKV waves then their defenses can be progressively worn down. This would begin the bombardment phase. The bombardment would likely last for millions of years with successive waves of beam and RKV swarms being exchanged between galaxies. Notice that the feedback time is the same as the distance between the galaxies – in our case millions of years. I.e. only millions of years later will the aggressor even be able to see what the effects of their first attack was. Until this point, the aggressor should likely just continue ‘firing blindly’ and dispatching successive waves of RKVs and attacking increasingly remote beam targets.

Once the defender defenses have been reduced sufficiently to enable direct invasion, invasion ships will be dispatched and decellerate. These will likely look like small von-neumann probes with the goal of making it through sections of the disrupted sensor and defense network of the defender to find niches in which they can self-replicate and grow. Many of these regions will likely be repurposing the materials of previously destroyed infrastructure of the defender which would have a large concentration of metals and other key elements. Once established inside, the attacking von-neumann probes should gather more real-time information about the state of the defender galaxy and send out new probes to potentially undefended or highly damaged regions. Eventually, they should aim to build sufficient power generation capability to create their own beam weapons which could potentially be devastating given the vastly less time lag when fighting from within the defenders galaxy – i.e. on the order of hundreds or thousands of years vs millions. Over time, the goal would be to build a self-replicating and hard-to-completely eradicate ‘infection’ of the defenders galaxy which can attack targets randomly and with minimal light lag. These can then coordinate with incoming RKV waves to soften up and damage defensive installations prior to RKVs getting into range. Over time (millions of years) the goal would be to slowly overwhelm the defenders from both bombardment and internal attacks and eventually shift the balance of power towards the attacker from inside the defender’s galaxy.

Generally the goal of the attacker is to inflict sufficient damage in the surprise first strike to significantly reduce the defender’s capabilities and resistance to successive waves of bombardment, thus over time wearing down the defender enough that the attacker’s von-neumann probes can start building up independent power-bases within the defenders galaxy with the goal of ultimately overwhelming them.

Defenders Doctrine: The defensive doctrine would look like the inverse of this. To minimize the damage from bombardment, the defender K3 should do the following:

• Avoid having any seriously predictable infrastructure over long time horizons. All large scale infrastructure should change orbits at least somewhat often (every few thousand years). A few random motions and the possible volume a long range bombardment would have to cover would rapidly become infeasible. If you are forced to have any fixed infrastructure this should be treated as sacrificial rather than the heart of your civilization. I.e. if you are performing energy extraction from the SMBH i.e. via the Moore-penrose process, you should not colocate all of your computational and mind infrastructure there as well. Instead beam the power to a bunch of distributed jupiter brains spread out throughout space on eccentric and unpredictable orbits which involve random orbit corrections.

Similarly, do not cluster civilization around Dyson spheres which are tethered to their underlying stars. Instead starlift out the hydrogen and distribute it to a huge fleet of mobile habitats for use as fusion power. These habitats can then shift on random and eccentric orbits and are vastly harder to hit through static bombardment.

Generally, this kind of mobility and spatial spread is somewhat expensive in terms of delta-v, lost efficiency of dyson sphere vs starlifting, and general communication issues of a bunch of mobile computational units vs having an extremely dense ‘core’, however it makes your civilization almost impossible to eradicate or even massively damage through distant bombardment.

Due to the light-speed limitations of communication, this will always be a challenging trade-off for K3 civilizations. Having a denser core civilization will enable significantly faster more powerful ‘minds’ to be created which must occupy dense regions to allow rapid communication between their subprocesses. However, these dense cores are extremely vulnerable to blanket bombardment. Conversely, ‘diffuse’ civilizations are extremely hard to inflict meaningful damage on from afar, but the total communication costs and time becomes challenging. Unfortunately, it is not as simple as being dense until you detect a threat and then scattering since intergalactic beam weapons can strike at the speed of light – i.e. without any possible warning and the infrastructure to build them is fairly dual use.

• Build highly redundant and backup industrial cores. Assume that some of your core centers will be taken out in the initial first-strike bombardment. Ensure that sufficient industrial capacity remains as a backup that doing so would not meaningfully impair your defensive and sensor capabilities.

Your core industrial capacity must ‘run hot’ and be highly visible from outside if it is doing anything, but it is still very feasible to have a large amount of ‘cold’ backup capacity which is placed deliberately onto obscure orbits which will make it very hard to target from afar and hit and destroy in the initial strike wave.

Similarly, ensure that all your fundamental sensing capabilities are both distributed, move highly randomly, and are redundant. Assume that the enemy has mapped out all the key infrastructure and patterns of your civilization in advance and will target these. Be prepared to have backups be scattered and stealthed and just be left alone for millions of years. Build a highly redundant sensor net so if some large fraction of your interferometers get taken out in the first attack you can still detect and fight off intrusions both of RKVs from subsequent bombardment waves and also von-neumann probes.

Similarly, ensure that mind-patterns are highly redundant and cannot be wiped out by chance hits on computational infrastructure such as jupiter brains.

Generally, a defensive and paranoid K3 will be highly spatially distributed, randomly, mobile, redundant and diffuse. A vulnerable/naive K3 will have huge concentrations of industry and mind in fixed small dense regions which will look incredibly bright to adversaries. A paranoid K3 galaxy will look almost entirely like a uniform mesh of waste heat with no obvious focal points of industry or energy generation and usage, or at least if there are such focal points that they move randomly over long time horizons. This trades off some energy efficiency and some speed and depth of thinking for vastly greater survivability in war scenarios. Attacking a paranoid K3 civilization will be like trying to punch holes in a fog.

• Build distributed, redundant, and omnipresent defensive sensor and laser networks. Fundamental to the defense against RKVs and invasion von-neumann probes will be a highly accurate tracking and interdiction system against any unusual masses moving into your galaxy. Maintaining a heavily redundant overlapping defensive system is of paramount importance in preventing huge RKV damage to key infrastructure and even second tier infrastructure, and secondly in preventing any enemy von-neumann replicators from being able to enter your galaxy in the first place. Ensuring that this net remains at high capacity even under heavy bombardment is extremely important. Similarly, you should expect that individual defensive installations might need to engage large numbers of targets in relatively short time windows, thus necessitating multiple installations with highly overlapping fields of fire. Similarly, these defenses need to be able to track and engage many converging relativistic projectiles with extremely sophisticated AI and electronic warfare capabilities and the system would need extensive testing in such scenarios.

• Maintain powerful second strike capabilities. Even if you get no warning of the first attack, you should be able to launch a powerful second strike bombardment on the attacker galaxy. This strike is likely going to be less powerful than the attackers since the attacker implicitly has millions of years of warning if they expect a second strike, but nevertheless, prolonged counter bombardment can also impose high costs on the attacker.

• Build strong enough internal fleets, lasers, and a sensor net capable of detecting and immediately neutralising any von-neumann probe that does manage to get through and start amassing resources in your galaxy. This will require detailed enough tracking of activity of any appreciable amount of mass or energy across the entire galaxy. This is relatively easy in peacetime but maintaining this under consistent bombardment must be ensured.

Generally, the goal of the defender is simply to survive the first bombardments with minimal damage (even if they come without warning) and then either second-strike in retaliation or just continue not spending energy on fighting while the aggressor galaxy spends large amounts of their energy on this campaign. If the defender can maintain and highly diffuse and mobile civilization, and does not suffer large losses from RKVs and the intergalactic beams then the energetic cost to defend is heavily in favour of the defender. This means that potentially once well alerted and/or paranoid from the start a K3 can survive this bombardment while only requiring the diversion of a relatively small fraction of its civilizational resources in response. This nevertheless comes at some steep costs, specifically being unable to ever concentrate a large amount of infrastructure in any fixed area, as well as, if there are any fixed elements still in the galaxy such as un-starlifted stars or the SMBH, rendering building infrastructure around these extremely dangerous indefinitely.

Since it is similarly hard for the defender to meaningfully attack the attacker if they also use take up a similar defensive posture, it is possible for ‘irrational’ attackers who don’t mind spending resources indefinitely just to annoy another civilization to maintain an essentially indefinite bombardment/siege of another K3 civilization. This generally seems to be unlikely since the benefits to doing so seem few compared to the costs which can be a small but nontrivial proportion of the K3s total energy budget, but is certainly possible depending on the utility functions of the attacker.

Overall, to sum up, while there are a number of extremely powerful attack vectors through which an aggressive K3 civilization could inflict severe damage upon a naive and unprepared defender with little to no warning, if the defender is highly prepared and has reconfigured their galaxy in such a way as to be ready for defense, it appears that most of the attacks can be defended against highly asymetrically – in that a defender must invest substantially less energy, effort etc into defending to thwart an attacker’s attack. Being under attack is certainly costly in other ways, and defenders will inevitably suffer losses – even large ones over time. However, on balance, it seems that the equilibrium favours the defender. The fundamental fact is that the attacker must cross massive distances in order to attack while the defender does not – this means that the attacker must expend large amounts of energy to accelerate and cross this distance and also that the attackers must operate with a substantial time lag (up to millions of years) and thus must act blindly substantially more than the defender has to. These factors tend to push the balance of power towards the defender. Moreover, the strength of these factors increase with greater distances. Here, we have considered perhaps one of the most favourable setups for the attacker – two galaxies that are extremely close and are within the same local group. If we are thinking about even larger scale warfare e.g. across galactic groups, then the timeframes balloon into tens or millions of years and at some point offense just becomes infeasible.

Conversely, space warfare is much less biased towards the defender on smaller scales. If neighbouring solar systems are fighting or if somehow a galaxy is split between warring factions, then offense and defense are much more finely balanced. Beam weapons become much more fearsome when operating with a targeting lag of years or centuries rather than millions of years, and similarly RKVs might be harder to defend against with less warning and with more ‘noise’ inside the galaxy rather than around its edge. Broadly, the same points hold however and the facts are that even at this scale, defense still has a small advantage if they utilize the same general strategies described here, just less of one.

Although this all seems hyper speculative, these kind of dynamics can be important for modelling the far future. Specifically, it implies that once a K3 civilization is established in a galaxy it is extremely hard to dislodge. This means it is likely that the universe will end up mostly cleanly partitioned between full galaxies of colonizers with perhaps only short wars early on in a galaxy’s lifetime to determine which colonizer faction will be victorious in galaxies where colonizers both reached it at roughly the same time. Beyond that, the fact that intergalactic warfare seems likely to be expensive, interminable, and bring little prospect of gain to the aggressor, makes it likely that such wars are rare in general and only pursued for ‘irrational’ purposes. For instance, once the colonization phase is complete, a pure paperclipper would be unlikely to invade neighbouring non-paperclipper galaxies since it would have to expend far more energy to do so than it would gain, and hence would end up with less paperclips overall than if it just paperclip-ified its own galaxy and then sat there for the rest of time. This also implies it is very important for early civilizations to colonize to maximize their share of the cosmic endowment because it is hard to fight back and claim inhabited territory if you are late to colonizing.

Another thing of importance to note is just how extremely vulnerable large fixed objects like planets or stars are in this universe. Anything on a predictable orbit can be sniped with impunity and complete surprise from millions of light years away. This means that humanity is super vulnerable in its early stages right now and also even when spread across multiple planets and building its first Dyson sphere. Any K3 or less civilization in the vicinity (if they see us) can directly destroy us at this point, which makes moving some sensible fraction of civilization away from obvious large orbiting bodies and into freely moving habitats far away from any obvious orbits an immediately important backup strategy. Unsurprisingly, it also means that we should invest heavily into trying to determine if there are any K2 or K3s out there in our intermediate neighbourhood (e.g. perhaps in the local group) since if so not only will they likely have colonized almost all galaxies we could reach first but also that as soon as they detect meaningful civilization around earth, then they could disintegrate it with extremely powerful beams with ease. In a variant of the anthropic argument, the fact that this has not happened yet provides some evidence against hostile local K3s existing. Although evidence of human civilization on earth has only really existed for about five thousand years (and even then we only started seriously emitting about 200 years ago), the general fact that Earth was widely inhabited with life has been visible for many hundreds of millions of years – recall that interferometers that can resolve with kilometer or less accuracy across intergalactic distances can be easily constructed by K3s. This accuracy is easily enough to resolve e.g. forests, coral reefs, grasslands and other features which clearly indicate huge concentrations of life, let alone just measuring coarse atmospheric concentrations of organics which we can do today with the SETI and any K3 can plausibly straightforwardly have for all planets in its local group if not further afield. Life in general across the universe is likely easy to spot for any seriously advanced civilization and our continued existence provides some evidence that either there are no aliens or if they are that they are not universally genocidal.