It feels lately like my decisions carry a lot more weight than they used to. So last night, I asked: what if I could game out different scenarios before making them?
I built a "Game of Life" simulation in Claude Code. A character is given a set of financial and lifestyle goals, then guided by a career coach, a life coach, and a financial advisor — with a fourth agent running the simulation itself. I gave the character some of my own skills and opportunities, but from there, he was free to make whatever decisions he needed to reach his goals.
As time passes, he has to make realistic decisions and live with realistic consequences. No avatar, no UI — just actions and consequences. The game pauses every six months of "in-game" time, or after a real step toward the goals, and hands me a summary. I can dig into a full report log to see every decision and result along the way.
The results were fascinating: marital stress, a family illness, career risks, real wins and real losses. I could also dial how punishing or forgiving life was — which I ended up needing, because the first run got bleak fast.
This is what I mean when I say I treat AI as a toolbelt, not a single-purpose tool. I didn't set out to build a planning product. I set out to see what happens when you point the same tools I use to ship apps and evaluate design systems at something completely different — a question with no clean interface and no obvious "right" build. That's the experiment: not "does this work," but "what happens if I try it."
Will I use it to plan my next move? No. But it was a genuinely interesting way to game out decisions and consequences I hadn't considered, and it gave me some things to think about. And it's one more experiment in the AI toolbelt.