My first RPG character was a smuggler.

I didn’t pick it because I understood balance, builds, or systems. I picked it because I saw the Millennium Falcon, and smugglers had blasters. That was enough logic at the time.

I also added gambling as a primary skill. I still don’t know why. It just felt like something Han Solo would have.

That character lasted three rounds before my entire party was dead.

But that failure taught me something I didn’t recognize until years later: how systems respond to chaos when you can’t predict the inputs.

What a Pen-and-Paper RPG Actually Was

When people hear "RPG" now, they think of video games. Stats, skills, builds, progression bars.

All of that language already existed back then in tabletop games.

The difference wasn’t the concepts. The difference was where the game lived.

A pen-and-paper RPG lived entirely in the room. You had a character written on paper are the stats, skills, equipment, damage. That sheet defined what you could reasonably do and how likely you were to succeed. There were no menus. No tooltips. No prompts telling you what was possible.

If you wanted to do something, you said it out loud: "I shoot." "I bluff." "I go right."

Then dice were rolled to introduce uncertainty. Everything else was judgment, memory, and agreement between the players and the Game Master. That was the loop: decision, roll, consequence.

The Game Master Built the World in Real Time One person ran the game. That was the Game Master (GM). They weren’t just enforcing rules. They were building the world as you moved through it.

You’d reach a junction. Left or right.

You choose right.

The GM describes a long hallway. At the end, a locked door. Maybe guards. Maybe a choke point. Suddenly a simple decision has shape, risk, and consequences.

Nothing was pre-rendered. The environment only existed because it was described and remembered. The world reacted to what you did, not because it was scripted, but because someone was actively responding to your choices.

In our group, the GM was my brother. By the third round, we were all dead.

Why This Matters for Building Systems

Looking back, running a TTRPG campaign is just systems architecture with dice. The GM manages state:

• What players know vs. what their characters know
• Faction relationships that persist across sessions
• Consequences that cascade from earlier decisions
• Continuity that spans weeks or months of real time

The system handles unpredictable inputs:

• Players don’t follow expected paths
• They combine mechanics in unexpected ways
• They make choices the GM didn’t plan for

The world adapts or breaks:

Rigid narratives fail when players go off-script Flexible systems respond and keep moving Good GMs improvise within constraints, not chaos

That’s the same problem every software system faces: users don’t behave the way you planned, and the system has to handle it anyway.

From Copying to Creating

My first playthrough wasn’t creative. It was imitation. I copied Han Solo because that’s what I knew. Smuggler. Blaster. Gambling. Those were my training wheels.

On later playthroughs, something shifted. I realized you weren’t supposed to recreate characters. You were supposed to invent them. Let them grow. Let them fail. Let them survive or die in ways that weren’t scripted. Sometimes the stupidest choices became the most memorable moments especially when the GM had no mercy.

Why I’m Revisiting This Now

I’ve been thinking about how to recreate that experience digitally. Not to replace the human element.

But to explore: can AI handle the GM role? Can it improvise, adapt, and build narrative worlds the way a human GM does?

Tools like Foundry VTT already exist for managing the mechanical side, dice rolls, character sheets, maps. But the real challenge isn’t mechanics. It’s procedural narrative generation that responds meaningfully to player choices. That’s what I’m experimenting with. Not as theory, but as an actual build.

I’ll be documenting the process on EngineeredAI what works, what fails, and how to solve for the gap between "AI generated a response" and "AI built a world that reacts."

If you want the full story of that first Star Wars game and what made TTRPGs work before everything went digital, read the complete post on HobbyEngineered.

The first time, I copied Han Solo. The next time, I used my imagination. This time, I’m seeing if AI can do what my brother did in the 90s: build a world that survives chaos without collapsing. We’ll see if it lasts longer than three rounds.