Yesterday evening, I handed a client request to AI and spent over 20 minutes refreshing the screen
The job was still running, but I kept switching back to the chat window, afraid I might miss an error. The waiting was draining: there was work to do, but I didn’t dare step away. Later, a friend showed me terminactory, a casual management game that runs directly in the terminal, where you build production lines and calculate profits. For someone like me who turns waiting into mental friction, it felt better than refreshing the page again.
What is it? Not an AI tool, but a “waiting room” for long AI jobs
Developer snipedbywifi built the game in Rust and ratatui for his own terminal environment in San Francisco. You can build factories, dismantle equipment, select regions, and check profits. There’s also a mock browser for browsing fictional forums and setting materials. The most work-relevant feature lets it receive Claude Code notifications: when AI sends a new reply, I can see the progress reminder inside the terminal game instead of guarding the chat window.
I got stuck on one step too: when I first configured Claude Code notifications, I assumed installing the game would deliver messages automatically. It didn’t. I reread the project documentation and found the hooks configuration. AI did participate in the early development, and the author didn’t hide that. Right now, it feels more like a small space to play for two minutes while waiting than software that improves productivity.
Your cost to replicate today: $0, about 20 minutes, and knowing how to use a terminal
The software is free, and using the ready-made version is enough; there’s no need to rebuild your own. The first step is to open the project’s GitLab page and download one of the available versions listed in the instructions. If you already have a terminal, just install and launch it as directed. The technical barrier is simple: being able to copy and paste commands and knowing whether your computer can run it. Estimated cost: $0 and under 20 minutes. If you’ve never used a terminal, the illustrated setup guide might add another 10 minutes.
My advice is to use it for “waiting anxiety” first, rather than researching every feature immediately. I’d run one AI job that takes more than 10 minutes and see whether the notifications feel natural. If setting up the command line sounds like too much hassle, this tool isn’t for everyone, and it’s fine not to try it now.
If I’m busy, I’d pick based on my stage
Just starting: If my work often involves waiting for AI-generated results, I’d try the original version for 30 minutes and focus only on notifications, so the game doesn’t become another time sink.
With 1–2 clients: If clients are pressing for urgent work but an AI job will take 15 minutes or more, I’d keep the game and my AI workspace open side by side, then set a rule to check results every 5–10 minutes instead of mistaking “waiting for completion” for active work.
Scaling up: If several people on my team are waiting on long-running jobs, I’d first check whether it can receive notifications from our existing AI coding or agent tools, then decide whether to recommend it. At this stage, it’s better suited to an individual waiting experience than treated as team monitoring or project-management software.