AI + Magic: The Gathering

Here's what I hated about Commander once everything went online: everyone just searches "best Atraxa deck" or "top Korvold combo" and copies whatever has the highest win rate. The entire creative soul of the game got replaced by optimization spreadsheets. People stopped asking "what do I want this deck to feel like?" and started asking "what's the strongest version of this card?"

I wanted to flip that completely. Instead of starting with a commander or a strategy, I start with the experience.

The whole AI-assisted build happened over about four days while Bailey was at a conference, which tells you everything you need to know about my ability to moderate a hyperfocus.

The part I didn't plan: the eight decks turned out to be a designed ecosystem. Every one answers a different question, and none of them overlap in either mechanical or emotional space. Politics, inevitability, pressure, and stories. I set out to build fun decks and accidentally built a set.

My wife Bailey and her friend Divya had never played Magic. They've played plenty of board games (Bailey's favorite game in the world is Everdell; we own something like six versions of it, all with tiny woodland creatures collecting tiny woodland resources). They love cute animals. They do not love reading 100 unique cards with eight lines of rules text each.

Here's the problem I keep seeing: people teach new players by throwing them into Commander. A hundred unique cards, three opponents with their own hundred, and a format that is called casual but is one of the most complex ways to play. Not understanding half the rules while everyone else combos off is the worst possible first experience.

So I built a five-night campaign instead. Bloomburrow starter kits as the base, custom proxy upgrades layered in over the course of the campaign, and a story where the mechanics are the plot. Bailey and Divya became the Embers of the Veil, defending their village from a spreading blight. Two heroes against me running the Blight, 2v1, so they could talk strategy together instead of feeling lost alone.

Magic: The Gathering has over 27,000 unique cards. Commander alone has thousands of viable strategies. The amount of content, data, deck techs, articles, and Reddit threads about this game is genuinely staggering. No human can process all of it.

But AI can. And when you point it at all that data with a prompt like "I want my friends to be laughing hysterically while I play a deck themed around woodland creatures committing crimes," it synthesizes everything and gives you options you'd never find by searching "best green/black deck." It's not replacing the creativity. It's supercharging it.

Same thing with the campaign. I described the experience I wanted Bailey and Divya to have, the games they already love, what overwhelms new players, and what makes Magic magical. The design that came back was progressive and story-first. I couldn't have built it that fast alone, and honestly, structuring it as a campaign might never have occurred to me.

The catch, and it's the same one from the deckbuilding project: AI is excellent at generating and terrible at remembering. Every good thing on this page came out of a loop where I generated fast and checked everything against Scryfall, a printed list, or a file on disk. The checking is not overhead. The checking is the work.