The game ended when commander stupidly died from a group of T1 units.

Running env:
* Spring 0.80.5.1
* RAI 0.601
* mod: BA 6.95
* map: 2Lakes

Situation:
* not enough energy to d-gun
* there is water nearby
* squad of enemy non-floating, non-underwater units destroying a commander

Fix:
* move into water (commander is amphibious unit)
* wait for help (constructor boats or underwater assistent) or go into safe area (according to threat map) or just sit there until health reclaimed automatically

The most stupid move a commander made is started to build water energy maker when too less health was left, then he just blowed-up.

Hope another AI developers prepare their AIs for this situation

Let me point you to few errors that your post contains:

1) You play outdated BA (current is 7.04)
2) You play vs AI on comends (it should be destroy everything, no AI is good at preserving their com).
3) You play old, ugly stupid ass map. Whats worse, its a complicated map, even for a human. AIs are to be tested on simple maps (comet, crossing 4 final, barren, DSD. Sands of war/Small supreme v2 for water).
4) You expect AI to execute some non-trivial tactics (non-important on global scale). Why isn't this trivial? -And what if enemy has torpedo weapons? -Commander is useless in water, and in this way his explosion at least takes out some of the attackers... etc etc.

Please, no offense! It's just its ok if AI can: expand, coordinate attacks, go to t2. Here i said :)

I don't care if it will be implemented or not. I just described a situation and i see it could be split into algorithm (if AI layer has access to discover where is water), and its implementation will improve a bit AI behavior. Personally I admire by AI smart actions during the game. Isn't it a goal of any AI maker to make his AI smarter?