I'm working on a space exploration game where humanity has become extinct and Earth is no longer inhabitable. A lone AI activates and its main directive is to find another world where humanity can be given a second chance (via cryogenized embryos).

The setting is one where space travel happens at sublight speed. Imagine some sort of interstellar Kerbal Space Program where you send out probes to other star systems. In the context of the game, a world similar to Earth is incredibly rare (which might very well be the actual truth) and we are potentially in a desolate corner of the galaxy (no aliens).

Assume the AI can function for thousands of years (that would also be an element of the game). At some point an Earthlike planet is found, but it is not a perfect copy of Earth: too cold, too hot, no water, too much water. Something's not ideal.

How could an AI weigh whether to focus on terraforming the found planet versus continuing the search?

Which terraformable characteristics would be addressable with realistic technology and hundreds or thousands of years of waiting? And which would be lost causes?

I imagine some planets would be unterraformable (or not worth the effort vs the results) without "type 3" technology (e.g. tidally locked planets, those without a magnetosphere, changing a planet's orbit...). Whereas maybe other actions (melting the polar caps or crashing comets to increase the water amount) might be more feasible?

But if terraforming is always feasible, then the search might end just after starting, or not start at all (just focus on Mars?). So I might have to rework the game idea.

  • 7
    $\begingroup$ Why not do both? Terraforming's going to require ISRU anyway,so use the robofacs you build there for terraforming to also build more interstellar probes that visit promising stars, and...it's Universal Paperclips all the way down. $\endgroup$
    – notovny
    Jul 28, 2020 at 22:13
  • 5
    $\begingroup$ Terraformation by some degree is about your only option. I think your probes would be hard-pressed finding a world that people can step around on right now without ill effects. $\endgroup$
    – BMF
    Jul 28, 2020 at 22:52
  • 2
    $\begingroup$ If you are interested in such stories, you could read "We are BOB" (Three books). One of the main subject, while not completely on this topic, is about finding planets for humans and terraforming them (but not much technical detail though) $\endgroup$ Jul 29, 2020 at 6:27
  • 3
    $\begingroup$ Love the Bob books. However, if the probes can terraform a planet, they why aren't they terraforming Earth? $\endgroup$
    – NomadMaker
    Jul 29, 2020 at 7:32
  • 2
    $\begingroup$ @TheWanderer in such a timeframe it would be reasonable to get "sustained" terraforming i.e. the oxygen rate and other factors are good while the technological intervention is ongoing, but if the terraforming stops, then the planet slowly reverts back to inhabitable or less-habitable state over a similar timescale. For example, Mars currently can't have a decent atmosphere pressure, if it had Earth-like atmosphere, then it would lose it over some thousands of years, however if some terraforming process can build up that atmosphere, then it can also continuously artificially replenish it. $\endgroup$
    – Peteris
    Jul 29, 2020 at 14:16

4 Answers 4


This isn't an Either-or Proposition.

Terraforming is going to be a long process, and will require using resources on the target planet to successfully complete. You'll need to build infrastructure dirtside to change the atmosphere, and place the biologicals you need to support Human life.

So instead of just searching for a single Earthlike planet, and putting all the eggs in one basket, the AI should be sending probes that are capable of building robotic factories that can build more probes, and capable of building the factories needed to start terraforming.

When the probe gets to a suitable target system, it can then both build more probes to throw to other potential targets, and the terraforming infrastructure. If the terraforming effort fails, or if the world was not as suitable as initially thought, just build more probefacs, and throw them to other candidate stars.

When the terraforming is reported complete on any of the worlds that are being processed, that world gets sent enough embryos to spawn a viable population of humanity from the nearest supply. Meanwhile, millions of other probefacs are continuing to scour the galaxy for Terraformable worlds, or resources to build more probefacs, expanding exponentially.

You did tell the AI to consider occupied worlds as non-terraformable, right?

Huh, something's been detected at the edge of the Solar System...

  • 2
    $\begingroup$ For this topic, look into von-Neumann-Probes for this purpose: en.wikipedia.org/wiki/Self-replicating_spacecraft $\endgroup$
    – Hobbamok
    Jul 29, 2020 at 12:50
  • $\begingroup$ "You did tell the AI to consider occupied worlds as non-terraformable, right?" - For the purposes of gameplay, I think it would be much more interesting to overlook this little fact (a la the aliens discovered in Gregory Benford's continuation of the Foundation series). $\endgroup$
    – Deacon
    Jul 29, 2020 at 17:36
  • $\begingroup$ This is basically the plot to the "We are Legion (We are Bob)" books mentioned in the comments on the original question. A lot of what you wrote here is covered in them, and with reasonably accurate physics involved too (plus only a little handwaving). I would recommend that OP read them before making the game, both to get an idea of what is involved and to prevent outright copying of the books' content. And because they are awesome books. $\endgroup$
    – techturtle
    Jul 29, 2020 at 18:10

Solar radiation

The amount of energy the planet receives from the star can be altered by moving it closer or further away in its orbit, however it is not something that a civilization smaller than K2 could easily do.

In the same way it would be very difficult to change the orbit in order to become less eccentric.


This makes the dream of terraforming Mars close to delusion. A planet with a surface gravity much greater or much less than the 9.81 m / s² that we have on Earth will affect all of biosphere.

In addition, even with a powerful magnetic field protecting from the solar wind, there is atmospheric escape with a low gravity; and this planets will already have an atmosphere poor in light elements that AI will have to import. In worlds with much greater gravity the atmosphere will have more pressure, requiring it to be less massive, and making the radiation protective layer above the surface much smaller.

Magnetic field

This is quite simple. Even NASA has a proposal for an artificial magnetic field on Mars.


This is a problem that appears for a terraforming of Venus that in a long-term project can be easily executed. The rotational energy of the Earth is 2.58 e+ 29 J while the rotational energy of Venus is 1.38 e+ 25 J. Applying more energy by dragging the atmosphere over the crust and bombarding lighter bodies would not be something very extraordinary

Axis inclination

The problem here is not exactly on a big or small slope, although a planet like Uranus will make things quite difficult. Having a more or less stable axis, unlike Mars, can be important in the long run.

Having a satellite big or enough to maintain balance can be attractive, although we don't know how common it is for a rocky planet to have a moon so big that it appears to be a double system. However, a large planet like Earth is also more stable than a smaller one like Mars.

Crust thickness

A very thick crust will prevent tectonism and AI will have to work in other ways to enable long cycles of carbon and other elements.


Both the mass and the composition must be taken into account. This part is relatively easy, unless the planet orbits a star very different from our Sun (which will somehow make the planet unviable), the composition of the other worlds of the system must have the missing volatiles. Extracting excesses from a very dense atmosphere on a suitable planet is perhaps a much more complicated activity than including.


Hydrosphere and lithosphere are worked together with the thickness of the crust and atmosphere. They don't offer much of a challenge then.


AI will certainly have in addition to human embryos all sorts of microorganisms, seeds and embryos in a food chain complete enough to support human needs, right?


Why not build large spaced based biospheres while searching for habitable worlds?

Your biggest problem is that it will take time to find suitable words regardless of whether they need Terra-forming or not. As pointed out by Rodolfo (above) any planet you consider for Terra-forming is going to have to meet some very specific criteria and the likelihood of finding suitable worlds close by is not high.

And since you've already stated that travel between stars occurs at 'conventional' sub light speeds this by default means that just reaching the nearest suitable stars is likely to take decades if not centuries. Add even more time for your probes to report back followed by more decades still while colonizing missions are sent and your mission completion time blows out enormously.

You can shave centuries of the recovery effort however by targeting local stars that don't have suitable planets but which are rich in minerals and elements needed for large scale space construction. With enough local resources its possible to build enormous (continent sized) rotating habitats. And you will many more star systems suited for habitat construction than you will ones with potential new 'Earths'. The best part being you can still build habitats while the search for Earth-like words continues.

Eventually of course you will finds suitable worlds but by the time you do you can have scores of habitats up and running housing hundreds of millions of people, plants and animals sharing information and technology and acting like stepping stones for the transfer of people and resources if required.

Clarification - as noted below in the comments; there's no reason the process wouldn't start in the Sol system first after all the engineering problems are resolved. However the machine may also consider that since we managed to nearly wipe ourselves out once before its odds of long term success increase if new populations are also located elsewhere outside the solar system.

  • 2
    $\begingroup$ Why bother traveling to another start? Put the habitats in the solar system. Even if Earth is a total write-off (though I can't imagine anything that would make Earth more difficult to repair than searching other stars + terraforming), you can still use it to anchor habitats (in orbit). $\endgroup$
    – NomadMaker
    Jul 29, 2020 at 7:33
  • $\begingroup$ Agreed, but the starting position was that the computer had decided extra-solar planets had to be located. Certainly the logical thing to do first is experiment with designs and construction techniques etc to iron out the bugs before you start building them elsewhere. $\endgroup$
    – Mon
    Jul 29, 2020 at 9:40
  • $\begingroup$ Couldn't everything happen? Some go to the stars, some stay. $\endgroup$
    – NomadMaker
    Jul 29, 2020 at 16:49
  • $\begingroup$ @NomadMaker I can easily think of a dozen ways Earth would be more difficult to repair than performing a sub-light search for other habitable planets. All you really need is something that the searcher cannot understand and effectively work around that cannot be expected to spontaneously resolve itself. Of course, there's also the possibility of the Earth just being gone (maybe Galactus or Majin Buu showed up at some point in the past?). $\endgroup$ Jul 29, 2020 at 18:17
  • $\begingroup$ In the real world, I can't think of anything that would prevent building habitats for the new humanity in the solar system, except a continuing war. Exploring another system sublight is a huge task. You'd need a computer dumb enough to ignore the easy possibilities in the sol system, and smart enough to handle things light years away where we know very little. $\endgroup$
    – NomadMaker
    Jul 29, 2020 at 18:20

Sit on the Moon and wait for a few thousand years

Eventually plant life will take over and clear up the mess on Earth. The new Earth will be rich in oxygen and there's a good chance fish will have survived. Given that the AI is intelligent and has unlimited time, this is the least risky and least expensive strategy. If after, say, 100,000 years (1 million years?), Earth is still not clean, then it's worth trying something else, e.g.

Terraform the Moon

It's nearby and in the right zone. It just needs technology that we are already thinking about to make it habitable. If the Earth is unrecoverable then the AI can make the Moon into an ideal habitat and humans can live there forever.

  • 1
    $\begingroup$ Gravity is the problem, leave humanity on the moon, for even a few thousand years and the potential for them to go 'back home' is limited. Everyone will be adapted to living at 1/6 G. From their perspective it would be like trying to colonize a world with gravity 6 x normal - they couldn't do it. $\endgroup$
    – Mon
    Jul 29, 2020 at 22:58
  • $\begingroup$ @Mon - I'm talking about the OP's "cryogenized embryos". Version 1 - wait for a clean Earth and then take them back and unfreeze them or Version 2 - if Earth is never going to be suitable then first Terraform the moon and, when it is ready, unfreeze them and let them live there forever. Note: I've edited to make this clearer. $\endgroup$ Jul 29, 2020 at 23:11
  • $\begingroup$ Doable, if sperm,ova or embryos can be kept viable for that long. You would probably also want to do the same thing with as many plant/animal species as you could save as well. $\endgroup$
    – Mon
    Jul 29, 2020 at 23:20
  • $\begingroup$ @Mon - I agree with you. With regard to the viability, I was going by the OP's statement, " hundreds or thousands of years of waiting". $\endgroup$ Jul 30, 2020 at 1:21

You must log in to answer this question.

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