The videos and text will cover a lot of the same ideas, with the text just being a more concise way of conveying them. I don't presume to be an expert on the AI; these are just my thoughts based on personal experience (which is still pretty limited for campaign).
Daily Challenges
Video notes: first 1:15 is me talking about my general approach to dailies; 8:38 I notice there's no stock delay for selling and consider what I think about that; 11:52 to the end has a few more thoughts on the AI and power, among other things.
Quick summary: probably not much to be done here because the main behavior exploited because the player can get away with things mostly because he'll outpace the AI by too much to be punished.
- The main mechanic is to take on very large amounts of debt to make your stock easy for you to buy up, boost your stock price by buying shares of the AIs (and possibly going for lots of power in smaller games), then sell out of your stock before interest ticks and you repeat the process.
- This isn't a problem in multiplayer because other players will keep up well enough to punish you for your low stock price by buying you out.
- AIs, on the other hand, generally don't have enough cash or barely have enough cash at the pivotal moments. If the AI dogpiled cheap opponents or was even more aggressive against very cheap opponents, playing with debt might get too dangerous to work, but it would also probably make the game a lot less enjoyable for struggling or newer players.
- Some of the prices move in quite predictable patterns, with primary resources spiking early, tanking, then spiking again in a lot of games. Jaiwera has talked about this before, but the AI's tendency to focus on spot prices means that it will often not only overproduce some resources, but also neglect cheap inputs, causing the initial price spike. Basically, the AI's profit maximization leads to some oddities in the market and it tends to overcorrect. In-the-moment profit maximization is an easy-to-understand heuristic, though, and can be serviceable when the problems don't get compounded too much.
- There have been a series of changes that I assume dealt with the way the AI handles inputs that are paid from debt. For a while, the AI largely ignored power and allowed trivial power wins, but that was corrected. I'm not sure how things have shaken out now, but between the large shipping costs the AI is willing to incur in a lot of situations and the $500 power price in this video, I am concerned that the AI might not be paying that enough mind.
- The AI selling out of extra stock (i.e. past 5) to upgrade is interesting, but it seems like a bad idea when someone is buying into you actively (8:22 in the video).
Campaign
Video Notes: first 6ish minutes are mostly general thoughts on what I expect from a non-simmed run and how I think I might approach it; around 9:50 you can see that a scientific AI founded in a big iron patch but placed steel mills away from his base because ???; at 19:46 the AI mutinies a solar panel right when night is about to hit and doesn't even target the one that would disrupt superconductor (though an AI did do that earlier at 15:34); at 25:45 I try to figure out how I actually managed to win in week 1 and what mistakes the AI may have made
Video Notes: 9:03 virus on reasonably profitable steel that doesn't make a ton of sense (though there was an iron deficit and things eventually got awkward as a result); 10:35 the virus animation outlasts the virus itself; 13:12 a confusing circuit overload on my water pumps when there are potentially better targets, which I don't think is an isolated incident (i.e. I seem to remember the AI COing primary resource mines more often than makes sense); 18:20 power surge animation moving through tiles is neat, but it's a bit weird that by the time it reaches the last tile, the PS effect is almost over; 23:15 I'm asked to decide between patents, but can't see prices to help me make a decision; 26:45 confusing water claim by the scientist; 39:40 the AI continues to prioritize nuking highs over mediums, despite the latter being more effective in most cases; 44:25 the restart mission button shows up normally despite ironman being on, which makes it unclickable, thoughts on simmed vs. not-simmed and the campaign in general
- I think Zultar made some good points on stream a few days ago that touched on a lot of the issues, basically that AIs don't treat debt with enough respect in campaign and that their single-minded push for buying modules can cost them quite a bit long-term. It's also interesting how the AIs place an enormous priority on buying the cheapest module rather than maybe one that would keep their production more profitable, but that's not necessarily a mistake.
- The AIs ship some things truly absurd distances (26:45 of part 1).
- AIs seem to tend to found in ways that don't really account for starting prices. They'll tend to take more standard spots near upgrade resources rather than really trying to fully exploit expensive resources. There's no really good example of that in the video since no primary resource ever got to truly ridiculous starting levels, but in general, I've found that players could get a substantial head start by just going into those expensive resources heavily while the AI only barely competes. I do talk about this a bit around 5:45 of part 2, but the carbon in that circumstance isn't really a great example, and I didn't play around it quite correctly.
- I haven't paid enough attention to most auctions or perk selection to have any opinion on those. Financial Instruments auctions, however, seem to be systematically undervalued by the AIs. In my simmed AI mission playthroughs, there have been missions where I'm allowed to win FI for 64-72k on Sol 5-6 and I make that much back in cash from a single AI on the first interest tick. There are arguments to be made that the other players might have paid off some debt and that it was worth less to the AIs (because I had no debt and they'd also have to deal with interest), but I still think they likely should've been more aggressive in bidding for it. That said, they weren't that far off.