Stellaris Dev Diary #385 - AI Benchmarks

[Soranya]

Different people on the team have different skill sets, and we are continuing to prioritize performance and stability as number one and two.

The AI has been one of the biggest complaints after they, which is why resources that cannot effectively contribute to those (like me, for instance) can be freed to work on it.

Balance has not been a huge target for us since there have been more pressing concerns, but we're doing a balance patch next week to cover some of the most critical issues, like telepaths.

Tbh. maybe the AI system needs a rework - the current way how the AI "knows" what does what in the game seems to be sub optimal. (Like it does not seem to know that it can respecialice/replace buildings on a bsaic level and so on.)

 

[SeraphAscending]

Now for the AI; I think a big problem is that the AI is trying to build toward targets over time, instead of trying to optimize their current situation. In this update, I find myself deciding what to build based on how many Civilians I have.

I think that puts it really well.
Deciding factors should be "what do i lack?", "what do i have in surplus?" and "can i use civillians/maintenance drones for it or do i need to reshuffle?".

 

[Imp0815]

It's steadily militarist these days. -...- , but militarists still end up on top.

I think i know why.

Godspeed in fixing the game up!

 

[Tigertycoon]

Scaling goals are just a terrible way to deal with AI. Randomly placing buildings and districts only marginally increases the AI’s resources—but one thing needs to be clear: planets need specialization. There is no problem with a spiritualist empire having more planets focused on unity and fewer on science, but they need dedicated worlds, not random placement. If the AI would just build specialized planets, all problems would be solved. You can still have varying personalities—aggressive empires that create large fleets, and passive ones with high defenses.

There should be no goal based on game year, only on what the AI has access to in terms of resources and pops.

Let’s say I give the AI a lot of planets with thousands of pops in year one. It should use those to play as well as possible, regardless of the year. If it’s based on year, things like advanced starts don’t even make sense.

Yes, the AI should have personalities, but it shouldn’t be completely braindead. Instead of this random, chance-based system—albeit weighted by personality—it should actively leverage its traits. All too often, it makes stupid decisions because of a (yes, weighted but ultimately) random roll. Look at the Galactic Community: first an AI supports a resolution, then tries to repel it—not because its situation changed, but just because…

The whole system needs to be rethought from the ground up.

 

[Jidovka]

Guys. I love your game. I played it for many years. But pls listen to one simple thing.

If my game lags too much with 10 ai empires, I will just set it to 5 or 4. That in itself will reduce the difficulty dramatically.

I'm playing this game for the story, for the rp, I'm playing it too see the content from different mods I installed... I never played this game for the AI empires and I don't want to. They are, as you said, just a background for the story. To limit my exploraion, to have an early batlle with, or to just do some weird stuff. And later in the midgame they are just... nonexistent, because I'm actively annihilating them to reduce ingame lag.

This is not normal.
I don't want to have even more reasons to annihilate them. It's weird. it's stupid. Your game already have things like crisis (mid and late game) to challenge the player.

Please, just make this game less laggy. This is literally all I want.
You promised that this update would be all about optimization. And I really believed it and waited for the new patch to come. But now all I see is that instead of pop units that wanted to change their job every day, I now have the same number of pop groups with their own separate growth, characteristics, etc and all of them want to migrate somewhere. I honestly think that this is a major part of the endgame lag now.

Because let's say you have three different races from different empires on your planet. It already can make 9 groups (workers, spec and elites). Then we can set it that each of these empires had different ideologies. Like one was militaristic spiritualist, the second was materialistic egalitarian xenophilic and the third one is xenophobic authoritarian.

It would literally spawn every possible faction inside of your empire (I actually had it in one of my previous runs after conquering two of my neighbors at the very begining). And the thing is that your citizens can change their ideology. So in a worst case scenario it would just multiply your 9 groups into some cosmic number of 60+. And it would add even more, each time you get a new race on your planet. As I said, each of these groups have it's own growth, characteristics and they all want to migrate (so they are constantly checking all of your planets for the free place).
Even without migration it's a massive calculation dump. On 3.14 I mostly played with a slow growth rate so it can actualy be a bigger raw pop groups number than i had pop units before. And the problem is that you don't even need your planet to grow into a lategame colony to have this effect.
If this is actually a reason for lag issues, than all of the xenophilic empires should be exterminated, lol. I actually checked it in my current run now and yes - even before the midgame with 7 total normal-ai empires, the xenophilic one has 30+ pop groups on average on their planets. And this is a situation where 1 of these other 7 empires is a hivemind, and 2 are xenophobic.
Like it's less than it was before if we take it as 1 group = 1 old pop. But it can scale rapidly if you have a lot of xenophilic empires in your galaxy.

 

[haloraptor]

have noticed the ai is back to making loads of habitats and now citadels(and not very good placed ones either) any plans on adding an ai limit for these again?

I conquered fifteen systems from an AI empire a few days ago and every single one had a citadel on it. I didn't even think they could build that many!

 

[ashbery76]

Bye Stellaris for at least 3 months.

 

[Sneaky_minon]

Its being said to death but AIs can keep their personality and still be significantly improved. They make lots of dumb decisions right now that are neither effective nor in character. A spiritualist could focus on unity generation and go to war more readily against a materalist just fine while making its planetary and empire wide economy more effective. This is the area that needs the most improvement. The AI can't use expansion as its only economic tool, it needs to be taught/programmed how to refine its existing economy.

As for difficulty modifiers: A lot of players never consider that most events and story chains are Player Only which means without any modifiers, the player is the one "cheating" a resource advantage. If the AI had access to AI-only events even half as often as the player does, that could help a lot. A dynamic way to do this could be creating some generic modifier-granting events and adding an on-trigger that events can call each time the player gets a permanent buff that runs for all the AI, 40% of the time giving them 1 buff, 20% of the time giving them 2 buffs, and 40% of the time none. This would help the AI scale alongside the player. Naturally, some AI would end off better than others, posing as bigger threats or more competent allies in the process, with the player at the upper end of the curve (since the AI would have 80% as many modifiers on-average).

Another option could be to scale the amount of AI personality's effect on economy and military with difficulty. At Ensign, AI personality is 100% in effect, but in GA, it might be at only 50 or even 20% of economic and military impact. The AI retains full opinion and galcom personality, but has a more humongous economy and military at grand admiral. This could even be a switch like tech costs, for players who want GA modifiers but a more roleplay focused experience, or a slider disjointed from difficulty entirely like AI Aggression. ("Economic Personality: Exagerated, Normal, Limited, Opinion Only)

I generally like this--except for the idea of having AI-buffs that trigger off the player getting something---then that would make getting some events that are nominally good a *BAD* thing because it buffs everyone else.

"oh man, I got the <blank> event which gives <blank> which is not very good--I guess I won't finish this chain"---> which is neither fun, nor lore/role-playing friendly.

 

[Space_void]

Because turning the AI economy into a black box that prints out numbers makes weird things happen if – for example – you conquer a planet, then lose it, then conquer it again.

I'm sure there would be ways around it. For example once a planet has entered the player's hands and re-enters AI hands remains as is until reconquered or the game ends. Instead it just contributes to the planet count for the AI.

My point is there are solutions that don't involve the AI playing a game its terrible at anyway.

 

[TreeCloudGrass]

AI will never settle further than 2 jumps from their border. Who controlled another side of wormhole, AI or you?

The other (galaxy) side was controlled by AI.

The whole cluster was empty, and the AI should have been lured by at least the 3 habitats in the entrance to the cluster. (which suddenly reminds me that more than 1 habitat in a system feels weird those patches...)

 

[durecellrabbit]

Honesty the AI before 4.0 was good enough, and that is the level I'd like it to return to.

Also agree with others that the AI needs to stop begging for vassalisation.

However if you are going to be setting benchmark numbers the AI needs to meet. The economy at the moment is pretty crazy, and it's unclear what is bugged, overpowered, going to be changed, .etc. So it's not like there can be clear guidelines on how strong the AI should be if it's not clear how strong we should be.

 

[Falrinn]

With vassasalization I'd say there needs to be a few critera:

Basically, they need to either be in such dire straights economically that giving up their independence in exchange for subsidies is the only way to avoid a default, or they need to believe their independence is under imminent threat so it's better to become a vassal of someone they actually like.

In the latter case they will also try to form or join a Federation before considering vassalization, and will be somewhat less picky about joining a Federation with people they don't like, provided they like at least some members enough. This will eliminate the weird situation where a empire will refuse to join a Federation but be willing to be pulled in via vassalization. Willingness to be vassalized by an empire will also translate to willingness to join a Federation the empire is part of.

 

[0th Law]

Echoing what's been said before, I think specialization is the main thing. Like... I think the best way to put it is that each planetary feature/modifier should also provide a weight towards a given planetary specialization. For instance, a dimensional portal should have a really large weight for Physics research, while a tidally locked planet should have a really large weight towards Energy generation. That sort of thing. Of course, those weights should be modified based on relative empire-wide balance of a given resource, so a huge energy surplus might make someone choose to specialize their planet for something else instead.

Crucially, though, once automation is unlocked, those specializations should then be pursued not based on population, but on power. That tidally locked planet only has 300 people on it? Doesn't matter, just slap an automation building in and build as many energy districts as possible, sans 1 of each other district type. The population will fill in later (and power districts can be changed out for city districts as needed depending on housing).

 

[Ranamar]

That's it? There's still no fix to maintenance drones? It's not even worth mentioning?

It's been talked about in detail in several threads here. Making the gestalt denizen the main amenities job renders gestalts a micro mess that's almost unplayable. Maintenance drones migrating away can crash your amenities, and there's zero way to have empty jobs on the same planet as maintenance drones, meaning you can't build jobs ahead of pop growth or as a target for automatic resettlement unless the planet doesn't need maintenance drones at all. So you can forget filling up a large world late game like a ring world or Rogue Servitor ecumenopolis (unless you're willing to manually resettle constantly).

Making logistic drones complex doesn't change that. They don't provide nearly enough amenities to rely on as the main amenities provider (also, they're trader jobs, not entertainers, so they certainly shouldn't be the main amenities job).

Plus, unless there's been changes not mentioned in patch notes, a fresh machine colony still has awful amenities issues due to zero proper amenity jobs, essentially requiring the immediate building of an amenities building just to be amenities-positive.

The fact that both building up brand new colonies and filling out late game large planets just sucks for gestalts, or at least for machine empires, has killed all desire I have to play this update.

Please, please, please prioritize a fix or at least acknowledge that it's an issue. Or if gestalts are supposed to play differently now, explain how gestalt amenities are supposed to work when maintenance drones can't be relied on.

This might be a bit rude of me to say, but it sounds to me like you already know the solution to your problem, which is to build one (1) housing/amenities building. You just don't like that answer because it's not what you used to do. (What I used to do was check the "manage amenities" box in automation and leave it to go, of course, and civilians don't quite do that.) I agree that logistics drones barely add enough amenities to take care of themselves, they aren't plentiful enough to do the old thing, and they aren't managed by automation the way the old "manage amenities" checkbox would handle them. I'm just saying that you clearly have figured out a solution, which happens to be the same one that I figured out.

 

[TeHb Z Shadow]

Here is my breakdown of my 2 games. I haven't finished them yet because of mid game lags, so they are on pause. Before 4.0 i played ~21 game (discovered the game 2 years ago at ~3.8.4).

Same settings for both games - GA, mid game scaling 2275-2350-2450 mid-late-end dates, 2-2.5x crisis, 1000 star galaxy, 15-30 empires, 3-6 fallen empires, no Advanced AI Starts, High Aggressiveness, zero broken gates (so no Gates at all), l-gates enabled, 2x wormholes, 0.25x worlds.
Next games will be with 2275-2325 mid-late dates and increase crisis up to 4-6x as i played in 3.14

Stellaris 4.0 saves for AI bench numbers - Google Drive

drive.google.com

1. Asari (Mass Effect) Evolutionary Predators, democratic, Egalitarian, Xenophile, Spiritualist with Diplomatic corps and Crowdscurcing (it was OP until recently but i didn't know that, so number will be lower, and because of science from factions i made less science worlds),
bioships.

• 2227 year: 3 colonies, 150 empire size, adaptive evolution +46/5k, main pops are 76% of total population
  • energy = 285 -1
  • minerals = 1.6k +50
  • food = 112 +61
  • alloys = 1.4k +17
  • cg = 983 -3
  • trade = 2k +30
  • unity = 3.4k +63 (5/5, 1/5, 2/5)
  • research = +228
  • fleet = 25/28
  • Megas = archive with 4/0/4 Specimen (Active 2/0/3)
• 2240 year: 5 colonies, 199 empire size, adaptive evolution +53/6k, main pops are 70% of total population
  • energy = 1k +29
  • minerals = 993 +49
  • food = 984 +159
  • alloys = 1.4k +21
  • cg 2.8k +14
  • trade 2k +7
  • unity = 6.9k +159 (5/5, 4/5, 2/5)
  • research = +289
  • fleet = 31/32
  • Megas = + my first Dyson swarm, archive with 7/2/7 Specimen (Active 4/2/3)
• 2298 year: 10 colonies, 466 empire size, adaptive evolution +628/22k, main pops are 41% of total population
  • energy = 19k +742
  • minerals = 22k +360
  • food = 28k +872
  • alloys = 2.8k +161
  • cg 73k +288
  • trade 105k +1k
  • unity = 28k +2.5k (6 x 5/5, 1/5)
  • research = +3.6k
  • fleet = 504/642
  • Megas = Full active archive (i am actively trading with AI for all their specimens), 2 ruined megas, 3 relics from Curators
• 2326 year: 30 colonies (13 were taken from purifier during current war, war started 5 years ago), 1134 empire size, adaptive evolution +1219/32k, main pops are 22% of total population (purifier race pops are 30% of total pops)
  • energy = 348k +734
  • minerals = 196k +2.2k
  • food = 393k +2.6k
  • alloys = 20k +1.1k
  • cg 217k -187
  • trade 405k +2k
  • unity = 542k +9.1k (full tree, total planet Ascensions 46 - 5-6-5-10-5-5-5-5)
  • research = +11.7k
  • fleet = 1501/1730
  • Megas = Full active archive, 2 ruined megas, 3 relics from Curators, Rubricator, Ether Drake Thropy, The Surveyour, Than's Throne. Mega Engineering will be researched in 10 month.

2. Post-Apocalyptic, Imperial Xenophobe, Militarist, Materialist, Warrior Culture and Relentress Industrialists. FE buildings from EE (storage included), Regular ships:

• 2215 year: 3 colonies, 103 empire size
  • energy = 1.5k +13
  • minerals = 659 +44
  • food = 6.4k +5
  • alloys = 483 +29
  • cg = 2.8k +11
  • trade = 765 +2
  • unity = 254 +75 (5/5, 1/5)
  • research = +114
  • fleet = 45/62 (empty corvettes for future merc enclave)
  • Megas = No archive, 1/0/1 Specimen
• 2240 year: 4 colonies, 161 empire size
  • energy = 2.4k +12
  • minerals = 2.1k +74
  • food = 6.9k +17
  • alloys = 490 +62
  • cg = 5k +24
  • trade = 5.4k +21
  • unity = 1.6k +102 (5/5, 5/5, 3/5)
  • research = +228
  • fleet = 97/122
  • Megas = No archive, 5/0/2
• 2301 year: 13 colonies, 613 empire size
  • energy = 119k +914
  • minerals = 75k +510
  • food = 217k +647
  • alloys = 4.4k +486
  • cg = 11k +306
  • trade = 126k +463
  • unity = 25k +2.7k (6 x 5/5, 0/5)
  • research = +5.3k
  • fleet = 397/549
  • Megas = archive with Full Specimens, 3 relics from Curators, The surveyor, Daedalus Seal
• 2367 year (1 AE Empire eats another federation, i patiently waiting for possible awakening 2nd empire and WiH, and making fleets): 25 colonies, 146 empire size
  • energy = 974k +3k
  • minerals = 338k -19
  • food = 273k +81
  • alloys = 227k +6.5k
  • cg = 267k +3
  • trade = 1.4M +5.5k
  • unity = 1.2M +8.5k (full tree, total planet Ascensions 32 - 10, 5, 10, 2, 5 )
  • research = +63k (many science Ecu <3 )
  • fleet = 6810/7075 (fortress Ecu rocks!)
  • Megas = archive with Full Specimens, 3 relics from Curators, The surveyor, Daedalus Seal, Ether Drake Trophy, Miniature Galaxy, Khan's Throne, Ethernal Throne. 2 Dyson Spheres, Sentry, Mega Shipyard, Science Nexus, Stategic Coordination Center, Interstellar Assembly

 

[Ludaire]

This might be a bit rude of me to say, but it sounds to me like you already know the solution to your problem, which is to build one (1) housing/amenities building. You just don't like that answer because it's not what you used to do. (What I used to do was check the "manage amenities" box in automation and leave it to go, of course, and civilians don't quite do that.) I agree that logistics drones barely add enough amenities to take care of themselves, they aren't plentiful enough to do the old thing, and they aren't managed by automation the way the old "manage amenities" checkbox would handle them. I'm just saying that you clearly have figured out a solution, which happens to be the same one that I figured out.

It's not an issue of the amount of amenities they give. They could quadruple the number of amenities provided by maintenance drones and you'd still have this issue. It's their behavior that's the issue. It's the fact that you can't have both empty jobs to be ready for pop growth/attract auto migration and have any pops providing amenities.

Housing buildings help the initial colony issue (though that's still stupid; you shouldn't need an extra building as gestalts just to be net positive on amenities). It can also cover amenities on smaller growing worlds to get around this issue, but I somehow doubt that a nearly full ring world or ecumenopolis is going to be covered by a few amenity buildings. And saying that every single world that isn't tiny should be using all its city district building slots on amenities and nothing else just so you can do basic things like building ahead and leveraging auto migration seems really silly. Especially when all it would take to fix this is the ability to tell maintenance drones not to promote into jobs or auto migrate if it would drop amenities into the negative.

 

[Ilab]

It is mind boggling that after removing trade routes, reworking pops etc (which where supposedly the majority of the performance problems), we still have bad performance, so now it is the AI the problem? Perhaps the problem was never properly identified? That is what this constant come and forth tell me. We removed things for something (performance) blindly?

All these things where supposed to improve performance significantly. Yet we still see the, 'oh, this new system is very bad for performance'. What AI levels are good? The same we have on 3.14. I see no reasons to mess much with that, specially considering the trend that everything that is touched lately gets worse. So, I would strive to have AI behave as they did before, or as close as possible.

But above all, I would prefer if the performance issues where properly identified and fixed instead of having a 'new guilty system' each week.

 

[Lenk]

I had to take a bit think about what to type here after reading the posts here on the forum. I am not being rude nor disrespectful. I love this game, its unique and there is nothing else out there like it. This game gotten me into modding and learning how to code things. It has been fun to play with friends over the years and playing fun games solo yet with 4.0 I think that fire has diminished. I felt like I been lied to, and I feel like I paid to be a beta tester.

Thinking hard about the content released with Biogenesis leaves to feel that it is unfinished in areas and makes me question some choices on things or feels outright lazy in other areas.

The Deep Space Citadel is virtually the removed Fortress mechanic from the earlier version of Stellaris but instead of being freely built it counts as a ‘megastructure’ with a limit per system. I am fine with those changes but what I don’t agree with is the idea of removing a mechanic only to add it back at a later date behind a expansion without little to no content directly tied to it kind puts a bad taste in my mouth. Because I bought the game when it was a thing and I practically paid twice for it when considering the expansion. You can polish something from before but don’t say its new when it already existed but with a new coat of paint.

The DSC also doesn’t seem to apply skins to it from the different mechanical shipsets, just the stark white hull no matter the shipset. Which goes against the norm for all the megastructures prior if I recall. I’m not expecting a unique model for each shipset given prior megastructures only applied a texture but not even doing skins for them for the mechanical shipsets just screams lazy to me.

The DSC not having a bioship equivalent model also feels extremely lazy. When A bioship empire builds it looks so out of place. Bunch of flesh platforms around a white smooth metallic surface. Now I can understand it if another empire builds them and I built the defense platforms, were just using the structure that already existed, but building the DSC it should look like my empire built it. Also couple of megastructures such as the Arc Furnace don’t even have a bioship skin applied to them It made me stop playing with bioships after I seen that. Visually I like it when things meld together.

The new Fallen Empire Is a wonderful addition to the game but didn’t anyone consider adjusting bioship empires that go down the route of cosmogenesis to use the Bioships of the Hive Fallen Empire instead of the normal Fallen Empire ships?

On a side note, it feels strange seeing the Fallen Empires not using the advance government system that best showcases their ascension paths. It’s a flavor thing but small details do add up over time. I think this was lightly touched on the machine age if I recall right?

The 4.0 update itself. It has felt like I paid to be part of a beta test, this mixed with the stuff above makes it feel tad worse. We are 13 patches in and 4.0 doesn’t feel marginally better. Performance is all over the place. Sometimes are fine other times it randomly gets bad with seemingly no reason why. but with the AI being absolutely braindead and the idea of them being fixed with the possibility of performance to worsen, it makes me question what the point of all the changes was if performance is gonna get worse. 4.0 was said to be a good performance update and right now it doesn’t seem to be one at all. Is it better in some places yeah but I cannot say for certain until the AI is actually functional.

When referring to the update overall, granted this update was going to adjust content from all areas of the game. It was major changes and I’m not expecting the world, I expected some bugs or some exploit because it’s hard to nail down everything in development because were all human. (I hope). But to keep some changes like the reworked leader system for the sake of making leaders simple to use after everyone in the beta hated it was a poor decision. I am glad that system was ripped out and the old one put back in. I don’t know if the system that affected traits was reverted as well, the one whatever position your leader is in effects the outcome of traits of on level up. I do thing that works well thematically.

I really like how with the new planet system I can make very specialized planets and I like the idea of districts swaps or basic resources districts having more usage. Just a lot of things to mess around with. I like to see this expanded upon in the future.

For things addressing the AI. I like to see AI interacting with another Ais. I would like to know what the AI’s goals. Are they wanting to take over X space or want to build a federation with Y. I like to help or hinder these kinds of goals. Allows a bit of roleplay and story creation, but with how it stands now the AI did seemingly random stuff from an outside observer.

I like to have my galaxy be filled with AI to make it feel alive as much as possible. I will typically play on the largest galaxy (1k stars) with max AI. It fills up fast and the AI and I either come into conflict or diplomacy blooms. Usually conflict.

I like to see them actually build megastructures and go down their paths of ascension, in the 3.14 I would see them at least have 3 or 4 get to their ascension paths before the end of the game, and rarely ever see them build megastructures like the nexus. I like to see them do more of that.

 

[Sneaky_minon]

I think people are misjudging this pop-rework a bit (maybe).

pop rework made jobs more efficient, but at the same time mangified the number of jobs by a factor of a thousand and introduced alot of mechanics that increased complexity of things a fair bit (the job efficency multiplier, pop modifiers, etc).

that said, I think its pretty reasonable to assume that there's still a lot of optimization that this new system can be given that the old system had already long since reached its point of dimishing returns.

I could be wrong, maybe there isn't alot of room and there's perhaps a point that it needed more room to cook; but blame the hard-coded time limit on the DLC release. (between steam, the devs and the executives, I'd blame the exec's 99/100 times).

I do agree that ships need to be between 10 and 50 times more expensive and the naval cap dropped by that amount (if not more)....having large fleets should be hard.

A novel idea (as far as I know anyway):
I think we should do away with the fleet cap (or loosen it) and then make it so that starbases/planets spawn corvettes or something when attacked so long as they have the right buildings; the purpose being to make it so that military buildings have more impact both planet side and starbase side, and to make it so there's a strong bias away from lots of smalll fleets and towards larger fleets relatively speaking. Lastly make it so that FTL'd fleets from losses are more likely to be destroyed so that fewer micro-fleets are made---maybe also make it so that fleets that go FTL in the same system always join together as well.

Between this, the smaller overall navy's, the smaller number of fleets most players/ai's will tend towards, and it should drastically reduce the compute spent on pathing.

Lastly, I don't know how much is spent on particles during fleet combat (apparently the artifact missle launch consistently saturates the particle count limit even for small engagements, haven't validated that myself but I believe it)....but we could reduce the firerate of all weapons by a factor between 2 and 10, if not 50-100, especially the missles/torps and the strikecraft by a small amount too. (this will make alot fo weapons have firerates measured in shots/week or whatever, but I think its needed).

oh, and Make Titans Scary!

 

[Alnarra]

So as a disclaimer, I only play on Captain, I've gog 2200 Hours in the Game and usually play with 2x Crisis / All Crisis with the standard for pretty much everything else on Huge Galaxies, I also play at slowest generally so obviously that's going to seriously paint how I look at what the AI is doing and react to it.

1. It Doesn't have access to a number of resources that are inherent to the player including Anomalies, Rifts, Archeology Sites / etc. and this can all have some pretty detrimintal effects on it's overall economy. Let's not lie, how many of us have had our bacon saved economically because of a well timed Anomaly Reward / etc. Givign the AI greater access to these (Have it go through the conversation trees, use it's civics / ethics / origin / etc. to put weights on how it would answer a given anomly, that sort of thing.
2. The AI has no "Playstyle", that is to say while civics have SOME weight on AI decisions, it doesn't shape a unique playstyle.
   1. The AI has no concept of "Being done" with it's territory, it simply expands until it bumps into another empire. In the case of a human I have a very clear concept from the get go "This is a Tall Empire" or "This is a Wide Empire". These choices are shaped by a number of factors, but the most important of which is origin, civics, and the like. I propose the AI be given 3 playstyles that shape it's "Playstyle" and determine how much of the Galaxy it's comfortable expanding to until it's happy. IE How many planets does it actually need? This could be done with a simple calculation at game start of total number of stars in the galaxy and then it's got 3 possible mods.
      1. A Tall Playstyle (Like your Guardianship Civics) Only want 2.5% of All Starsystems, and after that will turn their attention inward. This could be heavily encouraged by civics like Guardian / etc. or between 5-7 Total Planets, or things like Megacorps
      2. A Wide Playstyle wants wants around 5% of All Starsystems or or between 15-20 Planets
      3. A Expansive empire will keep expanding until they hit a wall, this could be things like devouring Swarm, etc.
   2. You could use this to then further sort of govern weights on which way the AI wants to expand, putting value (like a player would) on a system with a planet, and then further on a system with a planet that matches the planet types it likes. Maybe for Egalitarian Geneis Ark it favors planets not of it's type more, that sort of thought process. After that the priorty goes on systems which would make up it's border and have the least number of entry points. My thought here is the AI can never really do human stategy, but we can give it weighting on systems it would like, and if it does have a cap on systems it will take that weight can actually mean something.
   3. You can then further use this to put weight on the Technology's it goes for. Right now I assume it just goes for random picks, but no human player would ever do that. We all know there are some techs that are magnitudes more important than others. If I'm playing with Genesis Arks, I'm going to take Epigentic Triggers the moment it comes up. Maybe an authoritarian or militarist AI will more heavily favor weapon and ship techs while a peaceful or egalitarian civ will prefer economic techs. Hell maybe there are some techs it literally does not pick. Maybe a master Crafters Civic AI will never pick a shield tech unless it has literally no other choice, but will always take artisian and armor techs if they show up.
   4. I think that the AI should have economic goals, but only for 3 Things:
      1. Tech, Unity, and Alloys (Or Food in the case of bioships) and should use the other economic pieces purely to balance though. All to often you'll stumble into an AI player that's just sitting on a horde of one thing while something else is absolutely train wrecking their economy. If I'm running a game, I only run with as much energy or consumer goods as I need. I'm going to sell every drop of food I don't need and use it to get things I do (Consumer Goods / Minerals etc.)
   5. In terms of general "Goals" I have a few hard goals in any given game
      1. By Year ~5-10 I should be sitting on 100-150 Minerals / Month and while it's perfectly fine to go above that, in general the goal should never be to below that if at all possible.
      2. By Year 25-30 I should have 300 Combined Tech / 100 Unity / Month give or take
      3. By Year 50 - 500-700 Tech and Picking up my ascension path.
3. This brings me to some other things, there are some sort of fundamental parts of the game that we as human players just know and honestly I would argue they should probably just better incorporated as something that's default rather than picking a time and a place. At my 3rd Tradition Tree I'm picking my ascension if I'm ascending, at my Fourth If I'm going to go a crisis path I'm picking it there. As soon as I have mega-structures rolling I will just take Galactic Wonders just regardless. I think if we're not going to make these sort of baseline things, than the AI should heavily favor that sort of play-style as well.
4. Stealth, and this is probably a bigger discussion for the use of Stealth Overall, but the AI in general will not touch the stuff outside of science ships, but I think we should really look at maybe putting Stealth for Military Ships behind the Subterfuge Tree, sprinkle some cloaking bonuses in there, perhaps make the finisher be that stealthed Military ships that are not detected can bypass FTL Inhibitors. Make the effect of the Detection Array not based on the number of Hyperlane's from the station but rather use the tech you guys are using for cosmic storms where it's detection is in a radius around it. This could also put a lot more value on things like the Sentry Array, maybe Stealthed ships don't show up even if you have full intel on someone unless you have a sentry array. In this way it forces the player to think about "Ok where do I want to put my sensor stations to make sure the AI can't just sneak past?" and adds value for the player on the Sentry Array. I think if you put it in the Subtrefuge tree rather than it being a component, just like science ships now it's just there for all military ships, it could make that tree far more valuable and fun to play with and counterplay against. Maybe have epionage missions to "Blind Detection Arrays" and the player is notified "Hey the AI blinded a detection array" and so you reposition fleets at soft spots to make sure an AI fleet doesn't uncloak on top of you. These sorts of things.
5. For whatever reason the AI appears to inexplicably favor hitting 20 Corvettes as soon as humanly possible, and trying to fill it's Naval Capacity regardless of civics during the early game, and often times this is to it's serious economic detrimant. Much like it's tendancy for Endless Expanse I think this too should be more shaped by it's civics. Unless I'm sitting next to a Determined Exterminator, if I'm a peaceful I'm probably not going to be dumping a ton of resources into ships, have the AI use it's own civic / etc. weighting to make similar choices. Let it take a look at the neighbors next door and put weights on what it builds based on what is around it.

A lot of this really comes down to having the use it's Civics / Origin / Ethics / Traits / etc. to give far more weight to how it plays. I feel like it could honestly make the math a little simpler if the AI just stops looking for new places to add to it's territory at a given point. Maybe if it's a wide authortiarian empire that wants a few more planets in it's equation to be happy that's when it starts claiming systems nearby, but otherwise? Better to beat someone into submission as a vassal than claim land that's only going to make your empire more unweildy. Even if it doesn't necessarily make the AI "Better" at the game, i do feel like it will give it more personality and sometimes skill can be shaped by that personality.

Anyway just some stray thoughts.