I agree about the rally points. We could use an overlay on the main map that highlights possible origin points for a rally point, and a set of arrows along the intended path. The rally point menu itself should be roll-down (from the top of the screen) and transparent so as not to obscure the main map too much, and the origin points and arrows should update depending on which RP you currently have selected. It would be nice to be able to type the first few letters of a rally point name to select it from the list, gives a reason to change the default RP name.
It is asking a bit much, but as soon as there's more than about six rally points it becomes a pain to figure out where each of them is in relation to where you're looking right now (an off-screen arrow would be helpful!) not to mention you have to scroll through that teeny little box to select 'em.
As far as the AI goes, there's some work needed with research strategy. The AI either spends too much time on supportive technologies such as defences, engines, larger hulls, logistics and miniaturisation or too much time on weapons. It needs to strike a balance according to the current situation. There's no point getting too far ahead of the competition in one area while lacking in another area.
For example, my strategy is to first get to the point where I can fit two weapons to a fighter. It's pretty economical to build ships with this configuration, and gives enough military rating for other races to leave me alone. At this point if I don't need more military I stop building fighters. Then, as I'm usually playing Terrans, I go for the integrity grid and fit one per heavy fighter, giving them an advantage even against fighters with better firepower. Next, I develop better sensors so that I can spot an incoming fleet. Yes, I know I could just build Eyes of the Universe, but I don't need such a ridiculous crutch. Ideally after that I go for medium hulls and better weaponry, and enough of a logistics score that I can fit three heavy fighters with a frigate.
The reason why I don't immediately research planetary invasion is because even if I could build enough transports and fill them quickly enough, chances are my heavy fighters (in squadrons of three) will be easily swept aside by enemy frigates. I need to have my own capital ships and enough logistics to fend off the inevitable counterattack.
After that, a little improvement in each area is necessary. If an enemy is using missile weapons, then I'm going to research missile defences, but I'm not going to use the spare space on each hull to spam early missile defences, it's a waste of production. Unless I'm desperate, I would prefer to use a pair of PD Combo modules on a capital ship, or Telepathic Defence if it's available.
The scramble for the end of the weapons tree is understandable but predictable. By the time an AI has researched one of those, I'm already likely to have the best possible defence against it. There's nothing wrong with abandoning that first weapons tree and, while catching up to that level of weaponry in another tree, using the old weapons. Sustained missile fire will burn out missile defences and damage ships better than primitive beam weapons. When the AI abruptly switches to using its new plasma-based weapons, suddenly I'm the one having to adapt.