In the next version, if you didn't see already:
* AI Guard Posts are now all only direct-only targeted. Thus player ships (especially longer-ranged ones) won't accidentally free the ships around guard posts, and also won't accidentally set off any of the nasty special abilities these sometimes have.
** Thanks to CogDissident for inspiring this change.
* Human ships no longer automatically open fire on low-power AI ships, though they do still put them in their targeting lists as normal (so that as soon as the AI ship comes out of low-power mode, they are able to fire on them just as quickly as before). This prevents long-range human ships from accidentally stirring up the guards of AI guard posts unless the player specifically orders them to. In general this will make the AI guards bumrush the player ships a lot less frequently than they were recently doing, while not losing the first-mover advantage of the human ships against the low-power AI ships.
** Thanks to CogDissident for inspiring this change.
And this is coming also in that version, though it isn't on the wiki yet:
* On difficulty 7 and up, AI ships now spread out 2x as much as before. AI ships that are waiting before going through a wormhole to a human planet now spread out 4x more diffusely to make them even harder for players to hit.
* The AI now uses the intel data it has about the relative strengths of planets (as human players do), when determining whether or not to send its ships through a wormhole they are waiting on. This is different from the prior method, which was partly randomized, partly based on an accumulated number of 200 ships, and partly based on having a lot of ships incoming.
** This is a far-reaching change to the emergent behavior, which will have many effects on the gameplay, some of which are likely to be unanticipated at this time. The general expected result is that the AI will not attack players with "trickles" of ships very often anymore, and will instead choose to build up before breaching. The other expected result is the AI acting more sensibly when its command station has been prematuraly destroyed or when ships are freed from a guard post.
** On lower difficulties (<5), the AI actually overestimates its strength 5x, leading it to make dumber decisions. On difficulties less than 6, it overestimates its strength by half, leading it to make occasional stupid decisions there. Both of these are examples of the intentionally-sometimes-off decisions that make the lower-level AIs easier while also making reasonable mistakes a human might.
** On difficulty 9 and up, the AI actually underestimates its strength by half, leading to it to have a greater tendency to wait to strike with overwhelming force.
** These changes should also make the AI more effective in defender mode.