I am curious how the enemy AI is avoiding clumping together. This is a pretty important to address in this type of game and I have been researching the topic recently and I really like how the AI behaves in Redeemer.

Based on my observations, I would say the game is just using a lot of different enemy states mixed with making some enemies aggressive and others passive (when there are a lot of enemies targeting the player). Also adding slight randomizations to variables such as move speed.

Though it's also totally possible the game is using something more sophisticated such as context steering. I would like to know how you guys managed to create such good enemy AI, because they work really well in open fields but also work really well in narrow and tight spaces (which is where I would suspect context steering to break).