I use AI in my work.
Some people will stop reading right there. They have already picked a camp. They have already decided what the word means. In their minds, AI means stolen work, fake art, lazy people, and endless slop.
That is easier than asking what actually happened.
It is also wrong.
There are people forming mobs around this subject. They hunt for artists who used AI somewhere in the process. It can be a writer, musician, programmer, filmmaker, or visual artist. The details do not seem to matter. Once the tool is discovered, the person becomes a target.
It is an old story with a new excuse.
Find the witch. Build the fire. Ask questions after the smoke clears.
I will not hide my tools to make that mob comfortable. I will explain what I do, show the work behind it, and let the finished work stand as evidence.
My stories did not begin with a prompt
I started playing Dungeons & Dragons when I was eleven years old. I became both a player and a Dungeon Master.
My characters began at ground level, like most characters do. Over time, their worlds grew. A street became a city. Cities became countries. Countries became worlds. Worlds became parts of solar systems, galaxies, universes, and parallel dimensions.
The characters grew with those worlds. Their view changed as their experience and authority changed. A person who has never stood on the front line of a war sees war differently from a soldier. A soldier sees it differently from a general. A general sees it differently from someone responsible for the fate of an entire world.
That way of thinking has stayed with me for decades.
The people in my books are built from real experiences, private lives, old role-playing stories, memories, desires, arguments, mistakes, and questions I carried long before generative AI existed.
AI did not invent Nyla, Tiffany, Catherine, Vanessa, or the Doll House.
It did not live the experiences behind them.
It did not decide what they mean.
I did.
I do not ask AI to decide my vision
I think of the character first. Then I ask myself whether every action, sentence, choice, and detail fits my vision.
If it does not, I reject it.
I do not let AI decide who a character is, what she believes, where she should go, or what should happen to her. AI can offer language, arrangements, questions, and possible solutions. Those are suggestions. They are not decisions.
The final decision is mine.
That difference matters.
An AI-generated product can come from one instruction followed by little or no human judgment. AI collaboration is a process in which a person remains responsible for the ideas, direction, selection, correction, continuity, and final work.
Those are not the same thing.
The work is in the choices
I go through my work over and over.
I read it aloud. I listen for what feels wrong. I begin again after every chapter because the whole book must still feel like one continuous story. A chapter can be approved and later reopened because I noticed a problem, remembered something important, or realized that an idea needed to appear earlier.
I often ask myself, “If I could do it over again, what would I change?”
Then I change it.
Sheer Dominion became a 495-page book only after far more material was considered, written, changed, rejected, restored, and rewritten. There were many passes and many abandoned ideas. I do not claim to have counted every one of them. The exact number is not the point.
The finished book is the receipt.
A careless generator cannot maintain that many connected decisions across hundreds of pages by itself. It cannot remember every rule, relationship, object, joke, promise, visual detail, and future consequence unless someone builds a system to preserve those things and keeps judging the result.
That is why I built KATE.
KATE remembers the whole project
KATE is my private production and memory system.
It contains the rules, laws, character records, timelines, approved language, source material, visual standards, decisions, and evidence behind the work. Some rules are extremely specific. I have even banned certain hex colors because they do not belong in the visual world I approved.
That may sound obsessive.
It is.
That is the point.
KATE separates a source from a suggestion, a suggestion from an approval, and an approval from a finished result. Something does not become true merely because an AI said it happened. The system is meant to preserve what was actually decided and expose what still needs proof.
The canonical library contains hundreds, and likely thousands, of pages of context. That is more than one normal AI conversation can reliably hold at once. It also belongs to me, not to one AI company.
When I move between tools or begin a new conversation, I should not have to teach the entire history again. The canonical library and secure memory layer carry that history forward.
The AI changes.
The authority does not.
The environmental cost is real
AI has an environmental cost.
The data centers behind it use electricity, water for cooling, physical land, and equipment that will eventually become waste. Pretending those concerns are fake would be dishonest. Government and energy researchers expect data-center electricity demand to keep growing, with AI helping drive that growth.
Using AI does not excuse me from caring about what it consumes.
I cannot control how every vendor powers its data centers. I can control how carelessly I use them.
KATE helps us reduce waste by getting closer to the correct result the first time. The system preserves our rules, decisions, approved language, images, and finished files locally. Once an answer or asset is correct, we keep it. We do not repeatedly ask an AI to rebuild the same thing because someone forgot where it was saved.
We reuse approved masters. We make smaller versions from those masters when a website or social platform needs a different size. We preserve research so it does not have to be gathered again. We carry context forward instead of burning through another long conversation to reteach the same history.
None of those choices makes an AI request free. They reduce unnecessary requests.
That distinction matters.
Efficiency is not only about money or speed. Every avoidable generation uses more computing power. KATE chooses the engine required for the task instead of automatically sending every job to the largest available model. Locally hosted models are used when they are capable of producing the required result. Frontier models are reserved for work that genuinely requires their greater ability.
The goal is not to use the biggest machine. It is to use the right machine once and preserve the correct result.
Every lost file can lead to another generation. Every fake receipt can cause an entire job to be repeated. Good records, careful routing, correct outputs, and local preservation are part of our environmental policy because wasted work also wastes resources.
We apply the same thinking to physical products. When a practical environmental option is available for something we produce, package, print, or sell, we choose it for that reason. We may not always be offered the perfect option, but environmental harm is part of the decision instead of something ignored after the sale.
When customers purchase through a checkout powered by Stripe, Black Cat Media has deliberately chosen to contribute part of its revenue through Stripe Climate. Those funds support developing carbon-removal technologies. We do not claim that this makes the company carbon neutral. It is a contribution, not a magic eraser.
The answer to AI’s environmental cost is not to deny the cost. It is to use the technology with purpose, avoid unnecessary work, preserve what has already been made, choose better production options when they exist, and return part of the revenue toward repairing the world that made the work possible.
Hex was not made with one magic sentence
Before KATE, there was Hex.
Hex grew through a long visual and technical history. Cindy Lu became the machine-spirit Ava. Ava led to different versions of the Hex avatar. Those experiments eventually helped shape KATE.
That history was not one prompt followed by a finished character.
It was a chain of human choices. I selected, rejected, refined, combined, and corrected each stage. Some versions failed. Some systems became too tangled. HexOS eventually collapsed partly because the avatar, console, name, colors, and larger system had become tied together in ways that could not survive cleanly.
KATE became an evolutionary restart.
Technology helped me bring the character to life. Technology did not decide why she existed.
Calling Hex “AI generated” erases the person who imagined her, directed her development, rejected the wrong versions, and understood what she was supposed to become.
“Stolen IP” is an accusation, not an explanation
The next claim usually arrives on schedule:
“All AI is built on stolen intellectual property.”
That sentence sounds final. It is not.
“Stolen” is a legal and moral verdict. A person making that charge should be able to identify what was taken, who owned it, how it was used, and which right was violated.
Those facts will not be the same for every company, model, dataset, or output.
Some AI developers may have obtained protected works illegally. Some systems may reproduce protected expression too closely. If that happened, examine the evidence and hold the responsible party accountable.
I am not defending piracy.
I am rejecting the idea that one possible violation proves that every model, every use, every output, and every artist using an AI tool is guilty of theft.
Even the United States Copyright Office does not reduce the issue to that slogan. Its report says that different uses of copyrighted material in AI training can lead to different legal results. The answer may depend on the works used, where they came from, the purpose of the model, its output controls, and its effect on the market. Some uses may qualify as fair use. Others may not.
That is not a blanket verdict. It is a fact-based question.
Copyright also does not give anyone ownership of every idea, concept, process, system, method, or principle contained in a work. It protects a creator’s original expression. That difference has been written into United States copyright law for decades.
Learning patterns from material is not automatically the same thing as owning or reproducing someone’s protected work. An output that copies protected expression should be examined as an output. It should not be declared stolen merely because software learned patterns from many examples.
AI is not a human mind, and I am not pretending it learns exactly as people do. But artists have always studied other art. Writers read books. Musicians hear music. Directors watch films. Programmers learn from existing systems. Culture develops because people encounter earlier work and create something new from what they understand.
The real questions are about copying, access, transformation, substitution, permission, and market harm. Those questions require evidence.
They cannot be answered by chanting “theft” at every person who touched an AI tool.
A writer using a commercial AI service also cannot personally inspect every file that may have entered that company’s training process. A filmmaker does not inspect every line of code inside visual-effects software. A musician does not trace the ownership history of every software dependency inside a digital studio.
That does not remove responsibility. It places responsibility where it belongs.
Judge the developer’s conduct using evidence about the developer. Judge the artist’s work using evidence about the artist and the finished work.
For Black Cat Media, examine our ideas, source material, approvals, corrections, transformations, records, and final products. If someone claims we copied a protected work, they should identify that work and show the copying.
A slogan is not provenance.
An accusation is not a receipt.
Is Pixar lazy?
Pixar uses computers, software, digital models, simulated lighting, rendering systems, and CGI to make movies.
Does that mean nobody at Pixar is an artist?
Should the writers, animators, designers, directors, editors, technicians, performers, and composers refuse payment because a computer was involved?
Should audiences call the movie slop while it earns millions of dollars?
Of course not.
The computer is part of a production process guided by human judgment. Technology does not erase the people making thousands of creative decisions.
The same reasoning should apply elsewhere.
A digital camera does not remove the photographer.
Editing software does not remove the filmmaker.
A synthesizer does not remove the musician.
Spell-check does not remove the writer.
A compiler does not remove the programmer.
AI does not automatically remove the artist.
Not every use is the same
Some AI work is lazy.
So is some human work.
There are bad books written without AI. There are bad songs played by real musicians. There are bad movies made by large human crews with enormous budgets.
The existence of bad work does not prove that the tool is incapable of supporting good work.
The question is not simply whether AI was used.
The questions are:
Who supplied the vision?
Who made the choices?
Who protected continuity?
Who rejected the wrong answers?
Who accepted responsibility for the finished work?
In my work, the answer is me.
AI helps me cross gaps in grammar, expression, memory, organization, and production. Those gaps should not prevent my ideas from existing. I know what I mean, even when I need help finding the clearest way to say it.
The ideas are mine.
The standards are mine.
The final judgment is mine.
I use the tools available to bring worlds and characters out of my mind and give them a form that can survive me.
If that makes me the witch the mob came to burn, they should understand something before they light the match:
I built the fireproof library.