8/6/2026
Shaders, Collision Logic & Autonomous AI Companions
Hey everyone,
Here is a breakdown of the logical progress and systems architecture I’ve been working on for Blackwater Chronicles:
Technical & Visual Progress
I made some solid headway with the 1-bit shaders today. They are nowhere near final, but the foundation is there. I also implemented an "X-ray" occlusion system so you can see the player character when they move behind walls. While fixing that, I overhauled the collision logic—you will no longer get wall-locked by small bumps, and this natively solved the logic for traversing up and down stairs.
Story & World Concept
We are building a deep sandbox RPG, but Act 1 is highly structured to get you into the world.
- The Blackwater Den:You start as a prisoner here. It is considered the center of "freedom," but it's ruled by a ruthless, bloated Trade Prince named Leviathan.
- Factions: The city hosts the main guild halls for the Assassin’s Guild, the Thieves Guild, and the Pirate Cartel.
- Branching Paths:How you escape the prison dictates your starting trajectory. Depending on your choices, you will align with different "mentors" (separate from standard companions), which shapes your early game.
Gameplay & Scope
While it is a sandbox, the design is currently in a very fluid state. To deliver the tight, systemic RPG experience I want, I am deliberately cutting feature bloat. For example, free-building structures anywhere in the overworld is a no-go. The focus is entirely on combat, systems, and party survival.
The Autonomous Companion System (The Core Mechanic)
This is the mechanic I am most excited about. You [b]do not[/b] directly control your companions in combat. You play your character in real-time, and your party members operate autonomously based on their role, skill level, and current needs. You influence their tactics and roles strictly through conversational dialogue.
To make this work, they operate on a mathematical utility AI rather than a basic script. When a ranged DPS evaluates the battlefield, they don't just shoot the closest thing. They calculate distance (safety) versus the value of the target.
I’ve split "Threat" into two categories: Real Threat(base DPS/Healers) and Artificial Threat (Tanks generating aggro/taunts).
Here is the exact equation the AI uses to select a target:
How it works:
- St(Target Score): The AI scores every enemy. The highest score gets attacked.
- D (Distance):Ensures ranged characters factor in their safety.
- Treal vs. TartReal threat is actual damage/healing output. Artificial threat is forced aggro.
- w1, w2, w3 (Tactical Weights): These multipliers shift based on the role you discuss with the companion.
- epsilon(L) (The Experience Margin):This is the crucial part. Low-level (noob) companions have a high error margin epsilon. They panic, make mistakes, and fall for Artificial Threat Tart easily. As their Level L increases, that error drops. Veteran companions do pristine math—they become immune to basic taunts and ruthlessly prioritize taking down theTreal targets, like squishy healers.
That is a lot for one day, but the systems are coming together.
For full transparency: I am architecting all of this using an AI super-stack pipeline. I am not hiding that. My goal is to build and ship this game, and orchestrating these AI tools is what makes a massive, systems-heavy RPG like this possible for a solo dev.
I'm jumping back into the code. Open to any questions, thoughts, or feedback!