Games to play
Some ideas land better when you can poke them. Each of these takes something about AI at work and lets you feel it go wrong, then put it right.
Agents of Mild Panic
An AI maturity skill tree you play.
What it does
Pick a class, from Engineering to Legal & People, and earn skill points by sitting at the computer in a tiny pixel room. Spend them tier by tier towards a capstone. Work too long and you burn out; rest too long and you drift into irrelevance. Things go wrong while you play, and only the skills you built can stop them.
What it teaches
- Maturity isn’t how many skills you’ve done. It’s whether they hold each other up.
- Foundations come before the shiny stuff: a tier only opens once enough sits underneath it.
- Skills you stop using slip, so keeping them going is part of the work.
Scaling Agents
An idle game about running AI agents at scale.
What it does
Deploy workflows into AWS, Google Cloud and Azure and watch the agents hand off work, read context and call tools. As you grow, incidents, approval requests and outages arrive faster. Controls cover a set number of agents, so the cover thins as you scale. Scenarios take one thing away, like the context platform or the gateways, and let you watch what slips before you switch it back on.
What it teaches
- One agent is a demo. A thousand is an operating model.
- Shared context and shared standards decide quality. Without them you get rework and lookalike agents.
- Controls that worked at fifty agents don’t cover five hundred, and trust follows whatever you left uncovered.
- Without governing agents, a person approving everything becomes the bottleneck.
Your progress stays with you
Both games save in your browser and nowhere else. If you find a build or a strategy you're proud of, or one that broke something it shouldn't have, send it over.
