Building With AI Without Handing Over the Wheel

Trying to Build Tools for the Way I Actually Create

One of the biggest shifts for me was realizing I didn’t want a generic AI assistant.

I wanted tools shaped around specific kinds of creative work.

That’s what led me to start building Custom GPTs for my own needs.

GIDEON was built for fiction storytelling and canon adherence.
DORIAN was built for nonfiction, fact-based, documentary-style storytelling.
IVY was built for brand-based content creation, web, marketing, campaign work, and client management.
ADA was built as the final pass — the editor I use before handing work off.

Each one came from a real problem.

Not “how do I make an AI product?”

But “how do I make this part of my workflow cleaner, sharper, and more repeatable without flattening the work?”

That distinction matters to me.

Because I don’t code.

Building standalone AI apps never felt like a realistic lane for me. I wasn’t coming at this from a software founder mindset. I was coming at it as a creative person with too many moving parts, too many ideas, and too many moments where I knew exactly what I wanted to make but didn’t yet have the infrastructure to hold it all together.

Custom GPTs felt like the first format that made sense.

Not because they were flashy.

Because they were usable.

Because they let me build systems around process, voice, memory, and structure.

The obvious follow-up question, of course, was: will people pay for Custom GPTs geared toward their creative needs?

So far, the answer has been: not really.

At least not in the way I hoped.

That part has been discouraging. There’s still a gap between what I think these systems can do for creative professionals and what most people currently understand or value about them. A lot of people still see AI as either a toy, a gimmick, or a threat. Very few are looking at it as disciplined creative infrastructure.

But even that disappointment taught me something useful.

Sometimes the market doesn’t immediately validate the thing you’re building. That doesn’t always mean the thing is wrong. Sometimes it just means you’re early. Sometimes it means the clearest proof has to come from your own work first.

For me, that proof became Seas of Mystery.

Can a Custom GPT Like GIDEON Actually Work?

That question moved from theory into practice once I started building out Seas of Mystery.

The original version was a podcast: 14 episodes, each centered on a real maritime mystery, phenomenon, or unexplained event. Same host. Guests, experts, listener interaction. It had the feel of a dramatized radio show, and I loved that version of it.

But then I started asking a larger question.

Could this become something serialized?

Could this become something with the scope, consistency, and long-range structure needed to feel like a real television property?

Not just a podcast concept. Not just a collection of stories. A true IP.

Something that could eventually be ready to pitch at the level of a streaming series.

That changed everything.

The process started simply enough. I’d bring an idea into ChatGPT and ask for an outline. Then I’d take that outline and expand it into scene beats. Then those scene beats would grow into scenes. Then I’d do a human pass and shape it into something I could actually believe in.

That part, on its own, is manageable.

What stopped being manageable was consistency.

Once you get beyond a few pages, fiction starts demanding receipts.

Characters have histories.
Locations have rules.
Ships have significance.
Objects carry meaning.
Storylines branch and reconnect.
A detail mentioned on page 12 can break something on page 58.
A decision in Episode 1 can collapse Episode 7 if you aren’t paying attention.

That was the moment I realized I didn’t just need an assistant.

I needed a system that could help me maintain continuity across 60+ pages, and eventually across an entire season.

One pilot episode became ten episodes.

Ten episodes became a massive body of supporting material.

That body of material became a world bible that now stretches to roughly 2,000 pages.

That’s where GIDEON really came from.

It wasn’t built as a novelty. It was built because I needed something that could help maintain adherence to the IP’s canon.

With the right modules uploaded, GIDEON can search the world bible, confirm or deny a choice, suggest corrections, flag contradictions, and help keep the story moving in the direction I originally intended. It doesn’t replace the creative judgment. It helps preserve it.

That’s the difference.

The system doesn’t decide what Seas of Mystery is.

I do.

The system helps me hold the line.

What AI Actually Helped Me Learn

One of the biggest misconceptions people have about using ChatGPT for creative work is that it somehow “does the work for you.”

That hasn’t been my experience.

What it did for me was much more useful than that.

It allowed me to grow into kinds of work I had never seriously imagined I could do.

It helped me learn creative writing at a deeper level.
It helped me understand showrunning.
It helped me think in terms of fictional systems and serialized storytelling.
It helped me build something that can potentially move across multiple forms of media.

Games.
Books.
Comics.
Film.
Serialized television.
Toys.

Not because the tool magically made me a master of those formats. But because it let me practice, iterate, test, revise, organize, and keep going at a scale that would have been much harder for me to manage alone.

That matters.

Because there’s a version of this story where I never attempt something this large at all. Where I assume I’m not the kind of person who gets to build a world at that scale. Where I stop at “interesting idea” instead of pushing all the way toward “finished system.”

AI helped close that gap.

Not by replacing skill.

By helping me develop it.

AI as Tool, Not Crutch

I’ve also learned something else through all of this: using AI well requires discipline.

It is very easy to let these tools become a crutch.

It is very easy to let them overgenerate, overexplain, flatten your voice, or hand you something that looks complete but has no real internal life.

That’s why I’ve become so focused on process.

Prompting matters.
Command sheets matter.
Repetitive task systems matter.
Rules matter.
Memory matters.
Voice constraints matter.

If you don’t build structure around the tool, the tool starts steering the work.

And that’s exactly what I don’t want.

The goal, at least for me, has never been to surrender authorship. The goal has been to create a workflow where I can move faster without losing the human center of the work.

That’s the real test.

Can tools like ChatGPT be used in a way that lets the creator maintain control and still deliver work that feels human-written?

I think the answer is yes.

But only if the human stays in charge.

Only if the system is built to support judgment instead of replacing it.

Only if you understand that clarity beats convenience.

Is Seas of Mystery Good Enough?

That’s the part I can’t answer yet.

Is Seas of Mystery good enough to be optioned and developed?

I honestly don’t know.

Probably not. At least not yet.

But that’s not really the point.

The point is that I learned how to do it.

I learned how to build a pilot.
I learned how to expand a season.
I learned how to maintain canon.
I learned how to create an IP framework with enough depth to support expansion.
I learned how to think beyond one medium.

That’s not nothing.

That’s years of growth compressed into a process I might never have entered if AI hadn’t given me a way in.

So even if this version of Seas of Mystery never gets picked up, never gets optioned, never gets developed, it still changed what I believe I’m capable of making.

And that, by itself, has value.

Where I’m At Now

I’m still figuring out where I belong in the broader “AI for creatives” space.

I’m still asking whether there’s a real market for the kinds of systems I want to build.

I’m still trying to bridge the gap between what I know these tools can do and what other creatives are ready to trust them with.

Some days that feels exciting.

Some days it feels defeating.

But I keep coming back to the same thing: when I use AI with structure, boundaries, and intent, it helps me build things that are bigger, clearer, and more ambitious than what I could hold in my head alone.

That’s worth paying attention to.

That’s worth continuing.

And for now, that’s enough.

Want to See What Seas of Mystery Has Become?

If you want to see how Seas of Mystery has grown from a silly dramatized “radio” show/podcast into a much larger IP and world build, here are a few places to start: