Real-Time Agent Path Finding

Summary:
This is an Early Prototype Build of a Real-Time AI Path Tracing Algorithm that I will be using for another project for enemy hordes and more. 
All of the Levels are procedurally generated with no Nav Baking whatsoever.

There are no NavMeshes or Baking. Everything in this demo utilises a combination of A* Path Finding, Environment Vertex Compilation, and different game object layers, allowing enemy objects to understand Obstacles, Multiple levels, Complex Environment Spaces with Depth, and the player's position. 

The Algorithm that works in the background actively processes the environment and manages how agents can react to the player's movement, where they always find the nearest or most efficient path towards the player.

How is this possible? What type of Witchcraft did I use to pull this off?
One day, I felt that the baking levels I created became Incredibly Tiresome and limiting my creative spirit.

I used Navmesh in my previous game, Phantom Path, where its implementation with procedural levels was atrocious.
While Nav Mesh, Octrees, or other Solutions to AI Path Finding these days are used well by developers today, baking levels just take too much time for me, especially for the project I'm making.

So I woke up and had a hard think...How could I make levels that were not bound to any type of limitation...Then something popped up in my head, all levels used Poly gons, triangles and quads so what if I could compile all of that and implement an algorithm such as A* (A Star) and be able to tell a bunch of agents that, that was a moveable space, hence I got to work.

About a few weeks later, I have my Path Finding Algorithm that can deal with levels with a lot of Dynamics. Granted, AI did help me to optimise and fix some problems I had, but as of right now, I just implemented an Algorithm that developers back in 2013 were struggling with.

I had to watch a lot of videos on Path Finding, even a bunch of GDC videos where Naughty Dog developers were struggling to process more than 8+ agents at a time...Well, Naughty Dog, I figured it out, and while my Enemy AI isn't completely intelligent, I cracked how I could process more than 8, 50, 100, and 200+  agents on screen all in real-time with no baking, with a few tricks here and there.

So that is how I did it, Yay.

Was AI used in this Prototype?
AI, such as Chat GPT and  GitHub Copilot, helped me address issues that would have taken a few days to resolve. So Yes.

Will this be available to the Public?
No, not yet, while I have a basic prototype out, I want to squeeze as much Performance, Accessibility and Versatility into this system.
For now, it works great with 2D Levels with Depth and obstacles, but 3D levels with interconnecting layers that overlap still have a problem where the Algorithm starts to cheat...Bugs, am I right?

When will I release the code publicly?
After my First Official Game on Steam and GOG.
For now, you can play this demo and check out the AI Pathfinding in action.
Note* This Project is not the game I'm actively developing; this is just a prototype to test this Path Finding Algorithm.

What do I need to run this Demo?
Recommended Specs:
Windows 10/11
CPU with 4+ Cores
GPU with at least 4GB of VRAM

Developer Specs:
CPU: AMD RYZEN 5 3600
RAM: 32GB DDR4
GPU: RTX 3060 12G