Anyways, to start things off, here are some AI ideas that I have:
1. PyromanAIac: An AI that's designed to serve as a training tool for players on how to deal with an enemy that loves to focus on researching and deploying the dreaded Flamer line of weapons. As such, its research would focus on researching and upgrading Flamers, researching Hover Propulsion, Cyborg Factories, and VTOL Propulsion (especially at higher difficulties), researching Engine upgrades, as well as Kinetic Armor upgrades for both vehicles and Cyborgs, and, to a lesser extent, but especially on harder difficulty settings, researching and upgrading Mortars (including Incendiary Mortars) and Sensors, and later Bombs (including Phosphor, Thermite, and Plasmite Bombs) and Howitzers (including Incendiary Howitzers), to help deal with those pesky Hardcrete structures that are so resistant to Flamer damage.
2. GuerillAI: An adaptable AI that would primarily focus on wearing down its enemies through attrition: aggressively expanding for more Oil Derricks, utilizing hit-and-run attacks with Cyborgs, Hovercraft, and VTOLs, assaulting bases and securing areas with more durable Super Cyborgs, and Half-Tracked and Tracked vehicles, and sometimes even using Cyborg Transports to insert some Cyborgs, including a couple Combat Engineers, behind enemy lines in an attempt to build a forward outpost for a sneak attack on the enemy's flank. Would typically start matches off by having a couple of its Trucks build Machinegun Towers next to external Oil Resources that are closer to the enemy's spawn point than to the AI's spawn point, as well as sending out attacking forces to search for and destroy any enemy Trucks or Derricks outside of the enemy's spawn point. Would focus mostly on researching and upgrading Machineguns and Cannons, though would also focus on researching and upgrading Mortars, Howitzers, and Sensors (for both siegework and defense), researching vehicle armor upgrades and new vehicle bodies (including Collective and NEXUS bodies, as well as the Wyvern and Dragon bodies, but excluding New Paradigm bodies), researching Repair Facilities (and building them away from the base for damaged units to repair after performing a hit-and-run attack), researching research upgrades such as Synaptic Link Data Analysis (since many higher-tiered technologies, including Hyper-Velocity Cannons, Howitzers, Railguns, Lasers, and Composite Alloys, require them as prerequisites), and, sometimes, researching and upgrading Flamers and AA (in case the enemy was deploying slower Half-Tracked or Tracked units, or VTOLs, respectively). Would even attempt to adapt well to a wide variety of enemy unit and structure types. For example, it would build less VTOLs if the enemy had strong AA, build less Cyborgs and more Half-Tracks if the enemy had strong Anti-Personnel weaponry, build less Half-Tracks and Tracks, and more Cyborgs if the enemy had strong Anti-Tank or All-Rounder weaponry, build more AA units and structures if the enemy had been attacking with VTOLs, build more Cyborgs and Hovercraft armed with Machineguns or Lasers if the enemy had been attacking with mostly Cyborg or Hover Flamers, and build less units armed with Machineguns or Lasers, and more units armed with Cannons or Railguns, as well as more Cyborg and Hover Flamers and Hover Mortar or Howitzer Teams, if the enemy had been attacking with mostly slower Half-Tracked and/or Tracked vehicles (since the AI would actually attempt to take advantage of its enemies' units' slower speed, as well as their weapons' longer or shorter range).
3. n00bot: A joke AI for those who just want to relax after a hard day's work of compstomping in skirmish games or playing online in multiplayer games. Not only does this AI focus on researching EVERYTHING on the tech tree, regardless of how useless they are, but this AI would actually build some of the most pointless buildings that you could think of, like Flamer Bunkers or Tank Traps, as well as some of the most ridiculous unit templates imaginable, like Bug Flamer Tracks, or Viper Plasma Cannon Hover. Could actually pose a serious challenge if used in, say, a 1 vs 7 compstomp match, but otherwise, would almost certainly lose out against your typical AI, including Semperfi, DyDo, or even the default NEXUS AI.
Also, I have a few ideas for people to consider implementing in future AIs in general:
1. Building forward outposts that consist of, at the very least, a Vehicle or Cyborg Factory: This would allow the AI to deploy units closer to the front lines, where Oil Derricks may be contested. May also allow the AI to prepare a sneak attack on an enemy's flank.
2. Building Repair Facilities closer to the front lines: A smart player would know that preserving his/her forces and keeping them repaired would not only save them power from building replacement forces, but would also increase the odds that they live long enough to grow in experience level. One of the best places to build these Repair Facilities is close to the front-lines, where damaged units can quickly go for repairs, and then quickly return to the battlefield once repaired.
3. Using Cyborg Transports to airlift Combat Engineers (along with some escorts) to various places: The Combat Engineers could then proceed to build either an artillery firebase and/or airbase on top of a normally inaccessible mountaintop, or a factory (and defenses) behind an enemy's base in order to prepare for a sneak attack. However, some way to designate optimal places for such airlifting operations would be needed. Right now, the AI programmer might need to manually designate ideal airlifting coordinates for each map, though it may be possible to add support for such for map editing programs in the future.
What kinds of things would you like to see in future AIs?
And for the record, I wasn't exactly thinking about running whenever the weapon was empty, but rather, I was thinking about running if an enemy unit just got too close (such as close to their weapon range), hence the phrase "kiting". Also, I never thought about cornering an enemy either, though I do agree in that it would probably be difficult to program. 