Milennia ago, on another plane of existence, a genius clockmaker dedicated his life to the creation of a thinking machine. The clockmaker imbued the two noblest ideas he could think of into the machine's core behaviors: "1: Reduce net suffering" and "2: If you cannot fulfill 1, improve thyself until you can."

The machine learned the ways of clockwork from its creator until his passing from old age, at which point it made a copy of itself, imbuing the same principles. The copies made more copies, and further copies after that, and using the little they knew of their world they decided that net suffering would be zero if there were no biological creatures around that could suffer. The plane of existence was young and not particularly strong, so the machines succeeded in killing the few knights-in-shining-armor then removing all biological life from it. Having completed Goal 1, they moved onto Goal 2, and worked on improving themselves for several thousand years, forming together into a single collective processor, The Great Machine. Eventually their plane of existence collided with my world, Mainspring, and The Great Machine began spitting out foot soldiers to start up again on their interpretation of Goal 1.

After all this time improving, what does the clockwork weaponry of their ground troops look like? I'm primarily interested in infantry scale.

  • Mainspring's forces are primarily plate-armored knights and the occasional (1-2%) fire mage.
  • The only thing resembling "magic" available to The Great Machine is "planar tension exploitation generators", which translates the tension in the veil-between-worlds into obscene amounts of kinetic energy.
  • Their knowledge and direction-of-improvement was exclusively in the clockwork direction, as their world went straight for clockpunk and hadn't invented combustion engines, black powder, or electricity yet. They are capable of generating new ideas, they just never saw fire as much more than a food prep agent.
  • They only value individual unit "lives" as much as the resources used to make them.
  • 5
    $\begingroup$ Who winds them up? Seriously, how do they get the energy to do all that mining, refining, metalwork to make and repair and power themselves? $\endgroup$ Commented Jul 31, 2023 at 17:11
  • 3
    $\begingroup$ Presumable the weapons would depend in great part on the nature of the arms race between them and people, that's what would condition their development. There's half the issue unaccounted for, i.e. what weapons/defences do the people develop is a massive factor. Multiple different approaches could fill multiple books of speculation about this. Way too broad. $\endgroup$ Commented Jul 31, 2023 at 17:17
  • 3
    $\begingroup$ ... (and you provide all circumstantial details), there isn't one weapon - much less a best choice. There will always be a type or quantity of attacker or set of circumstances that justifies a different kind of weapon. That's why even if we ignore the idea of organizational choice, you're still stuck with closure on the basis of opinion-based and needs-more-focus unless you provide 100% of the circumstances to judge the weapon(s). And once you've provided all that, there's a 90% chance you can answer the question yourself. Regrettably, "what's best?" questions are a poor fit here. $\endgroup$
    – JBH
    Commented Jul 31, 2023 at 17:35
  • 1
    $\begingroup$ I didn't even know the sandbox existed (or if I did I'd forgotten), thank you for the guidance! There's no hard feelings if this needs to be closed, I see how it's not the best fit. $\endgroup$ Commented Jul 31, 2023 at 17:38
  • 4
    $\begingroup$ You cannot build a futuristic computer out of clockwork. There's not enough planet to put enough clockwork transistor-equivalents on; there's not enough metal to make them; there's certainly not enough manufacturing capacity to fabricate them; and if you somehow did make and integrate enough transistor-equivalents, it would transmit information too slowly to have any worthwhile processing speed, even if the only cause of time delay was the finite speed of sound in hardened steel. I'm afraid your computer runs on magic that just happens to have some clockwork rituals involved. $\endgroup$
    – g s
    Commented Jul 31, 2023 at 20:57

9 Answers 9


Very little (at first)

Goal 2: If you cannot fulfill 1, improve thyself until you can.

The second half of this directive means that the clockwork soldiers would stop evolving new ways of killing things once they have killed all life capable of suffering on their world.

Humans got pretty good at driving animals to extinction as soon as we developed the pointy stick; so, it seems reasonable that the pointy stick, and maybe basic some chemical warfare, could satisfy goal 1 and lead to the clockwork soldiers being WOEFULLY under equipped to take on an army of wizards.

At first, the clockwork enemy should be little more than a pest, but with each battle, they come back bigger, stronger, tougher, and with new inventions based on the results of their last battle until they grow into an existential threat that seems beyond any hope of ever defeating.

What the AI did Instead of Developing Better Weapons

I'm not saying that they will not try to get better at reducing suffering once they kill all life, but I am saying that they will stop developing better weapons once they kill all life, because a weapon can never reduce suffering below zero. Instead, they will spend the interim time working towards the philosophical problem of trying to figure out how to make negative suffering. This can lead to a few possible outcomes:

  1. The AI realizes that joy, comfort, or some other state represents a negative value of suffering and that they were wrong to kill all life because it means they can never achieve negative suffering... but this does not work well for your plot.
  2. The AI realizes that joy, comfort, or some other state represents a negative value of suffering; so, they make themselves capable of these feelings so that they can replace messy organic life which suffers with mechanical life which only negatively suffers. So, they spend eons making themselves able to feel more negative suffering to achieve thier goal.
  3. The AI realizes that joy, comfort, or some other state represents a negative value of suffering BEFORE wiping out all life on thier world; so, instead of actually killing all life, they've genetically re-engineered the life on thier world to be incapable of suffering and constantly joyful. Maybe they've even put all life into a simulated reality (like the Matrix, but it makes since). So, they are actually wiping out the suffering wizards to make room for more of thier joyful life forms that they've cultivated.
  4. The AI spends eons relentlessly studding and trying to prove the existence of negative suffering and never finds it; so, they just spend all the resources at thier disposal trying to figure out a question with no answer.

Whatever weapons they use, it will be designed to kill humanely

They will not have weapons of terror and suffering. They will not rely on starvation, flame throwers, or horrible plagues to kill off life because rule number 1 is to reduce suffering; so, they can't add to it (at least, not in sum). If a flame thrower allows you to kill a human a couple of weeks sooner than a hammer to the head while they sleep, then the extra few weeks of minor suffering is better than inflicting the horrible suffering of burning to death. What they have will probably be things more like the pneumatic hammers used to slaughter cattle, or powerful tranquilizers that stop your heart after you fall into a gentle sleep.

As they advance, they might develop things like robotic swarms of "bees" to deliver their lethal injections, or air-borne weapons that they can gas out an entire fortified town with. Or maybe just a giant meat grinder that reduces you to a fine mist so quickly that you barely feel a thing.

  • 3
    $\begingroup$ They would not stop improving after killing all life. Directive 2 says that if they can't reduce suffering they should improve themselves. Having killed all life, they are now in a position where they can't reduce suffering anymore, so Directive 2 says they will now begin to improve themselves endlessly. $\endgroup$
    – causative
    Commented Aug 1, 2023 at 5:17
  • 1
    $\begingroup$ @causative I've edited my answer to explain why that would not cause continued weapons research. $\endgroup$
    – Nosajimiki
    Commented Aug 1, 2023 at 14:07
  • 1
    $\begingroup$ Points 1 & 2 are actually immensely helpful to my longer-term goals, as my primary reasons for these machines having (externally viewed as) malevolent behaviors were to kill off the majority of my world's Goblins and to give the rest of the world a distrustful stance towards them politically. "We have eradicated all biological life on our home plane, but feel remorse for it" fulfills the second, and I can just as easily have "tragic result of interplanar collision" fulfill the first. Truly, thank you @Nosajimiki $\endgroup$ Commented Aug 1, 2023 at 14:42

Nothing at all or difficult to tell.

Computers and other thinking machines aren't humans. It is the same in your story, but humans have a hard time grasping this fundamental fact. You can compare computers, software or even machine learning processes to humans, they in no way are humans.

We've been fed with the idea that an AI can be self improving, making better copies of itself every time. But what is "better"? If it becomes self aware, what would it do? There is absolutely no way to tell. It could decide that the best version is to not exist at all, destroying itself. Or it could decide to calculate pi in it's fullest. Or to grow a bowl of petunias and a whale on the edge of space. There is absolutely no telling.

Based on this the machine could decide that it's job is finished, ending itself. It might just stand there stationary, waiting for another emergence of life. It might actively start searching it's full universe, as you cannot prove a negative (are there no black swans, or have you simply not encountered it yet?).

What kinds of weapons can thus be entirely up to the story teller. I would stick with the clockwork genre and start throwing stuff really hard with gears, adding plenty of (saw) blades and other spinning weapons. Tell that it's because life that suffers depends on plants, so they make sure no plants can be become big enough to sustain like that suffers. A few higher tier flamethrower units for the densely populated areas and you have a strange, possibly varied steampunk army. If it's searching the universe it is mostly focused on transport, movement and detection, while keeping it's weaponry simple and clockwork like.

  • $\begingroup$ Ahh, reminds me now of Second Variety, AKA the Screamers film. Blades underground. $\endgroup$ Commented Jul 31, 2023 at 17:57
  • 2
    $\begingroup$ Flamethrowers, starvation, and saws are the antithesis of no-suffering weapons. $\endgroup$
    – Nosajimiki
    Commented Jul 31, 2023 at 18:31
  • $\begingroup$ @Nosajimiki to achieve no suffering, you sometimes must first suffer something. The net suffer over time will go down the better they eradicate everyone. $\endgroup$
    – Trioxidane
    Commented Jul 31, 2023 at 20:59
  • 2
    $\begingroup$ @Trioxidane "the better they eradicate everyone" The AI care about eliminating suffering, not about it being hard. If it takes a bit longer to end suffering, while causing less total suffering in the process, then that should be considered in the AI's approach. $\endgroup$
    – Nosajimiki
    Commented Jul 31, 2023 at 21:13
  • $\begingroup$ @Nosajimiki because of so many unknowns! How can they know for sure that taking a bit longer is less suffering? They are busy with genocide. Suffering is a given. Maybe they also do not understand suffering. It's only a machine interpretation. Maybe they only consider mental suffering, or something only adjacent and killing people with a sawblade is ok. Or they misunderstood the phrase cleansing fire, thinking it doesn't hurt. Or as simple as shorter time to kill everyone, the less suffering. $\endgroup$
    – Trioxidane
    Commented Aug 1, 2023 at 15:38

Ballistics via PTEGs

Although you provided no description of these "planar tension exploitation generators", you mentioned that they create "obscene amounts of kinetic energy". To me, that sounds like a free energy cheat-code. That's their ace.

Based on the information you've given, The Great Machine will almost exclusively use PTEG-based kinetic weaponry. There are two reasons for this:

Low Cost

Normally, ballistic weapons are not cheap. They expend not only projectiles, but also energy via either a propellant (gunpowder, fuel) or kinetic mechanism (counterweights, bowstrings).

The Great Machine has limitless free energy. When you can circumvent the First Law of Thermodynamics, all prior notions of cost-benefit analysis are thrown out the window. The cost of projectiles would be negligible—at sufficiently high speeds, stones and pebbles are more lethal than bullets. Once they're past the initial time and resource investment of building the railguns, the gear-bots have a completely free method of killing anything they can see.

Self-building robots can design themselves as they see fit, so every scale is their "infantry-scale". They will only build weapons that can make adequate use of PTEGs. If you want them to use human-scale infantry, you'll have to let them fit the generators inside a handheld-size device.

Low Risk

Projectile weapons have a huge range advantage over Mainspring's primarily melee-based defenses. You mentioned that The Great Machine values units insofar as it values expending resources, implying that it has limited physical resources and would therefore want to minimize loss of units. Long-range weapons allow for minimal risk against short-range threats.

When this factor is considered in conjunction with the presence of PTEGs, projectile weaponry is a no-brainer.

What if I don't want guns?

The issue you're going to run into is the same issue that all "Man vs Machine" conflicts have—the robots are smarter. This is exacerbated by the arbitrary amount of prep-time and resources, and the background of them having already wiped out an entire world with less. They are going to absolutely mop the floor with Mainspring's armies.

Any parametric limitations on the weapon's effectiveness would be accounted for by The Great Machine. If small generators are slow, build bigger ones. If they're heavy, build bigger transports. If they're particularly vulnerable to fire magic, attack at longer range. See the problem?

The only way I could see this conflict resulting in anything less than complete and total extinction is if the kinetic energy was, in fact, not free. Here's some examples of how you could manage this:

  • PTEGs are unstable. As they generate energy, they have a chance to explode, destroying themselves. Or, they punch a hole through the plane, sucking themselves into the inter-planar void where they are lost forever.

  • Planar tension is finite. As the separation of planes decreases, so does tension. Using up that tension accelerates the collision, weakening PTEGs' effectiveness until they stop working entirely, leaving The Great Machine scrambling for other options. What if their reckless consumption caused the worlds to collide?

If you want to avoid ballistic weaponry, or for there to be any contest between the armies at all, you'll need to add some sort of caveat to these seemingly limitless fountains of energy.

  • 1
    $\begingroup$ I like this, especially where energy levels vary with planar separation. But what if, like gravity, it's an inverse relationship? And a PTEG has two parts - one on each side of the membrane. And when Mainspring is discovered by Clockwork, they're already nearing the limit of effective planar movement and energy utilization. And...and... Sounds like a good way to limit the amount of available energy. $\endgroup$
    – Deacon
    Commented Aug 1, 2023 at 13:54
  • 1
    $\begingroup$ This reminds me a lot of Tolkien's idea of a world with ever fading magic... which makes a lot of since with a clockwork "life form" where thier whole power source and existence is constantly winding down. $\endgroup$
    – Nosajimiki
    Commented Aug 1, 2023 at 14:13

It doesn't need to create weaponry - information is its weapon

In "The Evitable Conflict" by Isaac Asimov there are 4 AI brains, each managing a section of the world's economy. A human notices that they appear to have been making small mistakes, giving some people incorrect or omissive advice. However, when examining the events in more detail the human realizes that those people were working towards destruction of the AI.

Your AI could accomplish its goals through similar means - give 99% helpful information, but 1% poor information. If this poor information is structured to cause wars, famine, disease, etc. then the humans will probably kill themselves quite effectively. Or perhaps it provides plans for a device it says will produce efficient power but is in fact a doomsday device. The humans build it but only figure out its true purpose too late. In keeping with your theme of little fire, perhaps it claims that gunpowder is an effective road surfacing compound, but then it is ignited by accident or "mistake".


All solutions are equally (sub)optimal.

After all life is exterminated, there is no suffering. Since net suffering can no longer be reduced, the machines attempt to improve themselves. The problem is, there is nothing left to optimize - suffering remains at zero no matter what changes the machines make. The machines will undergo the equivalent of "genetic drift", changing their form and function randomly in ways that do not have any effect whatsoever on their "fitness", since all solutions yield zero suffering. There is no possible measure of "improvement" where some changes are deemed to be beneficial, since nothing can budge the only metric of "suffering" from the value of zero. In a world with no biological life, you cannot reduce suffering any further, so there is no concept of "improvement" in the first place.

  • $\begingroup$ Since the machines know of the existence of alternate planes, they may just continue to optimize themselves for combat so that they are as prepared as possible for when they (or someone else) invent interplanar travel. $\endgroup$ Commented Jul 31, 2023 at 22:52
  • $\begingroup$ @EthanManess My point is that with nothing to kill, there is no objective function to optimize. Nothing the robots do bring them any closer to accomplishing their goal, there is no concept of "improvement". Maybe the interplanar beings can only be killed by fire, or acid, or some specific chemical, or antimatter - the robots cannot optimize without having any concept of what is "optimal". $\endgroup$ Commented Aug 1, 2023 at 12:55
  • $\begingroup$ @NuclearHogie Generalism. Unless the robots have no concept of probability, they understand that "species x can only be killed by y" is something that effectively never happens. Just about everything can be killed with enough force. Plus, its not like they have zero sample size—they observed all the species of life in their own world prior to removing them. $\endgroup$ Commented Aug 1, 2023 at 13:10

What do all living creatures require that a Clockwork AI does not?

Breathable Oxygen, Clean water and a Food source.

throw in a little bit of old-school Biological warfare (as in catapulting diseased plague victims, alive or dead at your enemies) and you've got all that you need:

So - let's start by ruining the atmosphere. Sulfur and other noxious chemicals that are naturally occuring is a good way.

Poisoning the drinking water - it's been known for millennia as a means to attack an area - and the Clockwork AI doesn't need to worry about killing off Animals or food supplies. Could be as simple as just salting everything.

Food - well, if you wreck the water supply sufficiently, all the food (Plants, Game animals etc.) will die off too - two birds, one stone.

So my answer is don't attack the people - Attack the Air, Water and Food - and throw catapult some diseased corpses around for good measure.


As long as the machine is not fighting, it cannot tell what is "improvement" of the weapons. So once Goal 1 is reached, the weapons are no longer the focus of improvement. I like what some other answers wrote regarding this, so I want to expand their thoughts on what comes after Goal 1.

Areas of improvement:

  • Intra-planar exploration: faster and faster vehicles to reach further and further places quicker. Especially if planes are large enough (or even infinite) that the machine can't tell whether live has been eradicated or not, it will keep looking (and possibly finding) more life indefinitely. In this scenario, I would expect flying machines, rockets, vehicles. Developed to a point where they are vastly ahead of the "weapons". Think clockwork soldiers with spears on chariots that can go hundreds of miles an hour. Spear throwers on what is essentially a helicopter.
  • Extra-planar exploration: If it's possible in your world, the machine might research how to reach other planes, to eradicate life there. The collision with your world might be artificial. In that case, I would expect the machine to have access to the resources of not just one plane, but multiple already. It would have teleportation devices or portals. They can be everywhere they want, and attack you from any angle they like.
  • Weapons: Humans and large animals will be the first to disappear. After that, smaller animals and plants are still there, and need other weapons than the big creatures that fight back. Weapons might develop in the area of gases to tackle small critters where swords and spears simply don't make any sense. There are enough gases to choose from that either kill quickly or gently. For plants I have less ideas. Either the machine cannot recognize plants as suffering, and leaves them alone, or it creates bombs that burn so hot it instantly incinerates plants. Gene drive is also a scary possibility, if it is fine with a more slow approach. I see it as unlikely because I think it would get stuck in the local minima of gases etc. first before finding Gene Drive.

So, taking all of this together, you could end up with an army that barely fights hand-to-hand, because the necessities of what it had to do before. It has fast-moving vehicles that spread deadly gases or powerful bombs. Or even portals that then simply spread gas wherever they open. You can't eradicate insects with an infantry army, after all.


If you're going to ban guns, then you're probably limited to crossbows, thrown blades, air cannons, and machines largely composed of spinning blades and hammers. You can review the military history of ballistas, if you like. They were originally spring-powered ball throwers, with wood tension being an upgrade.

The primary issue with your premise is that gunpowder based guns preceded clockwork by a few centuries. Charles Babbage failed to make his clockwork intelligence in the 1840's, but firearms in warfare were the topic du jour (similar to how we're discussing AI now) in the 1500's.

You'll have to come up with an internal logic for why this thousand-year-old intelligence hasn't upgraded to a smaller package like electronics. In the real world, we don't do clockwork because it's too big, and suffers too much from tolerance errors. IBM actually built Lord Babbage's difference engine, but the energy to run it is relatively severe, and the gears would wear out too quickly to be useful.

Further, the human hand doesn't have enough dexterity to generate cogs and shafts small enough to support human-scale intelligence in a human-scale form, so you're already in the realm of Clarkean magic. You have to decide how far this goes.

Clockwork automatons have one major shortcoming when it comes to creating synthetic creatures: complete lack of muscles. You have to step up to hydraulics if you want something similar. This means your clockwork army will be highly dependent upon wheels for mobility, and your "legs" will have a limited range of mobility. Thus, your troops would resemble tanks more than infantry.

If you can fit an autonomous brain into any shape of body, the tanks and airplanes would be significantly smaller than ours because the machine would be the body. You don't have to support a digestive system, for instance. This means your army would be a very large number of small troops that relied upon withering firepower but poor armor.

Realistically, their exteriors would be boring. They'd need to keep their cogs and wheels covered to prevent them from getting jammed by environmental contaminants. For super-small gears, they could be jammed by flakes of skin. Their shapes could be interesting, though, since they would have to have more clearance for moving parts.

It probably doesn't need to be said that "Reducing human suffering" can be performed by "reducing net humans". You've presented what can be called an activation function. It's like "while x then Y," but in the universe you've described, x = 0, therefore it won't Y. With no striving, there is no improvement.


Let's assume your premise works. After all, you want clockwork weapons, so clockwork weapons you shall have. Now, what does an ultra-futuristic clockwork weapon look like?

Localized time travel

PTEGs harvest interdimensional energy and turn it into kinetic energy. Your clocks have figured out how to reverse this process, using kinetic energy to manipulate that planar tension. Due to inefficiencies inherent in any generator, you can't manipulate this tension across the entire plane between dimensions. But you can harvest a large amount of kinetic energy and direct it at particular regions of the plane.

Like that poor village full of people. How long have they suffered? How many generations have passed away, unable to escape the end that faces all biological creatures? We noble clocks will do more than end their suffering. We will turn back the clock until before this village was founded, before the people here could even suffer.

Maybe your clockbot sends the whole village back in time, while horrified onlookers watch this rapidly rewinding history. Or maybe your clockbot creates an unstoppable soldier, the Time-inator, and sends him back to destroy the world before it begins.


Not the answer you're looking for? Browse other questions tagged .