Dev Log 12 - Presents Under the State Tree

Hello my friends. It’s that time of the year again. The holiday season.

To some, this means hot cocoa, gifts, family dinners, decorations, joy and cheer. To others, it means congested traffic, financial stress, and pressure to be a consumer and purchase presents to show friends and loved ones that you’re a decent human being that cares about them. To most, I suspect it’s some mixture of these, and it is my hope and wish that this season contains more positivity and happiness for you than urgency and stress.

The rollout of Patreon Exclusive content has been steady, as can be seen in the update log in the Vault. Of course, December is one of those months that stretches into five weeks rather than four, and, being the holiday season and end of the year, I’ve treated it like four anyway, spreading out the content release of the last week across weeks 4 and 5.

I’m starting 2026 with a new course, as most of you have likely seen the spoiler. State Tree is a pretty highly-requested topic, and one I’ve been very excited to dive into for some time now. Many are debating the future of Unreal Engine’s suite of AI tools (game AI, not LLM AI), and whether State Tree will replace Behavior Trees altogether.

From what I have gathered, I don’t necessarily think so. It feels to me more like an additional tool that can be used where appropriate. Eliminating Behavior Trees altogether from the engine feels like an unlikely change, and even if that were the long-term goal, it would likely not happen for some time (I wouldn’t be surprised to see that Unreal Engine 6 still contains Behavior Trees, once we cross that bridge).

That being said, State Tree does seem to feel like a more polished and efficient solution for many things Behavior Tree could do (and many things it can’t). I personally like the feel and workflow of it more than Behavior Tree. I can readily perceive its potential performance gains and efficiency improvements based on its architectural design alone.

One of the more exciting aspects of it, to me at least, is the fact that it’s a bit more general-use than the Behavior Tree system. For those who aren’t well-versed in Behavior Tree, it’s an Unreal Engine system (really an AI architecture pattern that existed outside of Unreal originally) for driving AI Pawn behavior. It’s well-tailored to gameplay designers, with its own Blueprint-like graph where you can construct a Tree with nodes designed for architecting NPC behavior. It has its limitations; for example, Unreal’s implementation is tightly integrated with AI Controllers and Pawns.

State Tree, on the other hand, is much more flexible. While a State Tree can run on an AI Controller and access said controller for AI-specific purposes, this is by no means a requirement. State Trees can be placed on a simple Actor and used for any scenario that would call for a State Machine. I won’t dive too deep into the technical details in this dev log, but the way a State Tree transitions between states is fundamentally different from Behavior Trees in that a State’s Transitions can be optimized to be event-driven and ticking can be kept to a minimum, so State Tree offers a (potentially) more efficient and performant solution than Behavior Tree in general, and State Tree need not be used for AI at all. A puzzle game, for instance, with various states corresponding to stages of progress, could be driven by a State Tree with no ticking behavior at all, but rather transitions that await decisions made by the player-controlled character. All of this is, of course, dependent on project conditions and design, and of course State Tree can have ticking Transitions and States, as needed. And Behavior Trees do support some event-like behavior, despite the fact that they are layered atop a ticking system. At any rate, they are two systems with their own strengths, and for general NPC AI, Epic appears to be opting more for the State Tree as the go-to choice, as evidenced by their use in the new template projects.

So the State Tree course will get you started with the system, and we’ll be building up all the concepts throughout the course, seeing potential applications, both NPC AI and non-NPC AI alike.

In other news, I’ve got something exciting to talk about. The new State Tree course is being recorded in 4k! That’s right, videos are 3840x2160 pixels. None of my previous tutorial series lectures have been recorded in 4k, so this is a first, and I’m trying it out to see how much of an improvement this makes to viewer experience. If this is something that you value, feel free to comment on that.

I’m already planning the next course after State Tree, and this one requires some theoretical prerequisite knowledge. I know, I’m being vague, but it’s still early, and I can’t reveal all my secrets in one go, so expect more news on the next project in the very near future.

As the year winds down, I’ve had some life things to deal with, not least of which is the holiday season itself, so progress has needed to slow a bit, as seen by the spreading out of December’s 4th week across weeks 4 and 5. This is always a busy time of the year for me and tends to be less productive from a work standpoint, and while it’s a nice break and opportunity to visit with family and friends, it can tend to be a source of pressure and anxiety for me as work is sitting there, staring at me, and I feel its tug. As we begin the new year, I look forward to diving back in full-swing, and keeping the stream of content flowing. Patreon membership is on the incline, and it is my aim to continue that trend, and you, my friend, are helping to contribute to that reality. As it further materializes, more and more exciting things will continue to be possible, and for that, you have my utmost gratitude.

Happy Holidays,

Stephen