Welcome to the HexCrawl Dev Blog
Hi, I’m Paul Siegel — a software engineer and developer with 25 years in the video game industry. While my roots are firmly on the engineering side, I’ve always gravitated toward working closely with design and art teams. Getting to wear multiple hats is one of my favorite parts of the job.
Twenty-five years in, my résumé is full of companies that don’t exist anymore. That’s the industry - studios close, projects get shelved, and work you poured years into can disappear along with them. So alongside the day job I build things that are mine, on nights and weekends, where I get to decide what gets made and what it stands for. Right now that’s HexCrawl.
Why This Game
I’m an avid TTRPG player and co-host of Wandering DMs, a podcast about D&D and tabletop roleplaying games. I’ve spent a lot of time thinking about what makes those experiences special — the sense that the world is reacting to you specifically, that your choices matter, that no two stories are alike.
HexCrawl is my attempt to bring those feelings together: a cozy, tactical hex RPG where your character’s personal goals and history actually shape the world around them. Every adventure is procedurally generated, but it’s not random noise — the AI working quietly under the hood is reading your character’s story and using it to make the world feel like it knows you.
Why AI (And Why I’m Trying to Do It Right)
I’ve been chasing this particular problem for a while. On the Wandering DMs channel I ran a series exploring whether AI could serve as a D&D dungeon master — not as a gimmick, but in a way that genuinely replicated the feel of a responsive, creative human at the table. That experiment taught me a lot, and I wanted to take those lessons somewhere more controlled.
HexCrawl is that somewhere. The AI isn’t generating the game — the game is deterministic, handcrafted where it counts. The AI works in the margins, personalizing your adventure so it feels authored rather than assembled. I care about this distinction. I know AI is a loaded topic right now and I don’t take that lightly.
The practical version of that: real art wherever I can get it, generation only where it does something hand-authoring genuinely can’t, and a lot of ongoing work to make the inference small and cheap enough that it isn’t wasteful. I want to be as up-front about my AI use as possible, and help others understand the nuance of the choices I’m making, so most of what’s here is me measuring it, analyizing it, and often getting it wrong in public.
A Bit More About Me
Outside of day-job game development, I’ve shipped a few indie projects over the years. Most recently I created Fearful Ends, a horror pen and paper RPG that I successfully funded on Kickstarter and is still available online. I also created a mobile game called Road of Kings inspired by the solo board game Barbarian Prince about a decade ago. The influences of that game on HexCrawl should be very readily apparent.
Where to Start
This blog is where I share progress, talk through design decisions, and occasionally vent about the parts that aren’t working yet. If you’re new here, these four are the ones I’d hand you first:
- Reification — a player types “I carry a letter addressed to a woman named Elara,” and the game has to make both the letter and Elara actually exist. This is the clearest example of what I’m really building.
- The Glimmerwings Problem — the LLM invented an enemy I had no stats and no art for, and fixing it changed how I structure every prompt in the game.
- Writing the AI Out of Barbarian Prince — the time another project started with “let’s make AI do it” and landed on “that’s not the right job for AI”
- Groq Model Speeds — what the generation actually costs, in cents and in seconds, with the numbers.
You can play the current build in your browser right now, though fair warning: it’s pre-alpha, and it shows. If you’d rather see what else I’m up to, paulmsiegel.com is my home base for everything I build, HexCrawl and otherwise.
— Paul