While this post is not about my experiences in playing 3.0.3, I hope it will be helpful in finding the culprit of the issues.
I've spectated a lot of games in 3.0.*, both modded and unmodded and, I think, I can pinpoint many reasons for possible failures of the ai. This is a list of things that (according to my impression) are responsible for 70%+ of the causes of the breakdown. Also it's important not to believe me without checking, I do not have access to inner workings of the code and I don't know how most of the systems work in detail, my assessment might be wrong.
/common/ai_budget
I think mineral_budget spending on planets should have large static_min past the early game (at least 400). So whenever AI wants to build mining district it will always be able to, AI sometimes keeps spending veeery low mineral income on stations instead of increasing said mineral income.
/common/buildings
Main problem here is usage of two scripted triggers, "buildings_simple_allow" and "building_relaxed_basic_income_check". Both of them seems to be written pre-3.0 (they are located in /common/scripted_triggers) so they fit pre 3.0 economy better, each of them assumes that districts should be prioritised before buildings. In pre 3.0 economy buildings were responsible for specialist jobs and construction of them could crash the economy, this is why these triggers exist.
This is fine for some buildings, but is absolutely harmful for amenity buildings. Each of them requires buildings_simple_allow and this trigger will wait till the districts are built. This stops the construction of holo-theaters, low amenities crash the economy which further stops the trigger from evaluating to yes. AI keeps building more districts, amenities go lower and lower and economy never recovers. At the very least this trigger should be removed from the holo-theater, but it would be better if they were removed altogether (and replaced by a building-specific triggers)
/common/districts
Right now the field (ai_resource_production) makes AI think that an industry district will always provide consumer_goods. This isn't true on forge planets, right now all industry districts show as if they give 1 alloy and 1 consumer_good. It would probably be better if instead industry districts on the planet with forge designation had ai_resource_production of 2 alloys, industry districts on planets with factory designation showed ai_resource_production of 2 consumer_goods and planets without designation showed 1/1 split.
Second, industry districts probably need a trigger in allow = {} similar to "building_relaxed_basic_income_check". While it doesn't happen in 3.0.3 (more on that later, in /economic_plans section), in 3.0.2. AI was regularly overbuilding those tryign to follow the current economic_plan. If there was a trigger checking that energy,food. minerals income are greater than 0,0,50 respectively this wouldn't have happened.
Also from my experience if you increase ai_resource_production of rural districts (mining, farming, generator), AI is more likely to take seriosu care of raw resoruce deficits.
/common/decisions
Population control and discouraging growth shouldn't be used, same with "halt robot assembly" (if AI truly wants to halt robot assembly, it's better to have a destroy ={} field in the robot assembly instead).
/common/edicts
Somewhat of a surprising suggestion, but low difficulty AI really can't afford nutritional plentitude. It usually causes only starvation. The edict should have much lower weight (or, honestly, weight = 0).
/common/colony_types
Honestly, even though visually everything looks terrible when observing (none of the designation make sense, the horror), I don't think think this is major. Overall I think bureaucratic designation is overused and the best way to make AI do "correct" designation is to have a scaling weight. If the weight of each designation (mining, farming, bureaucratic, research) was increased by 3 for 1 building/district, by 6 for for 2 buildings/districts, by 9 for 3 buildings district the resulting distribution would be better and more pleasant for humans to see (more mining districts = proportionally higher weight). This way as a result of a weight comparison designation will be chosen depending on what was the largest amoutn of relevant buildings/districts.
/common/economic_plans
As I see, 3.0.3 economic_plans were a big change for 3.0.2 in that alloys and research income/focuses were removed. This does indeed help with 3.0.2 problem of AI overbuilding industry, but it has a side-effect of AI not prioritising science and alloys nearly enough (it's frequent to see AIs with 15 alloys per month in 2400 according to some of the screenshots in this thread). To fix it is possible to keep 3.0.3 economic_plans, but add a subplan that in case of good raw resource production (potential = {minerals, energy, food > 100,20,5 } for organics and 100,20,0 for non-organics) adds some research and alloys targets.
Alternatively it's possible to instead come back closer to 3.0.2 plans and write a trigger that disallows construction of new industry districts if economy situation is dire (or, actually it's possible to do both. Extra redundancy doesn't hurt the chaotic system).
/common/starbase_modules; /common/starbase_types
Hydroponics is a very very good module, it can really help low difficulty AIs if it was allowed to build them everywhere (and had high weight of it). While it sounds awfully specific, but it indeed a large change in how ensign AI can manage its food requirements.
/common/species_rights
Slavers will rebel. I don't think anything can be done with it (AI can't resettle, even human slaving empires can't be managed without resettling), apart from disabling AI slavery which isn't a good solution. Some things can help though, for example I think any slaver should prioritise identured servitude over any other slavery type (except for serviles, serviles can have caste). I know, it's not #optimal in human case, but this would lower risk of rebellions massively.
/common/technology
Some technologies need to be almost autopicked by AI, namely economy technologies (I suggest to increase the weight of these technologies at least in 10 times, preferably 100). I am talking about technologies which give 20% boosts to miners/technicians/farmers, technology that unlock hydroponics and technologies that unlock resource edicts (capacity_overload, mining_subsidies, etc.).
Frankly, I don't know whether these are "optimal" or popular among players, but they immensely help low difficulty AI manage their upkeep and production.
.png)
