What to do while your AI codes.
You hit enter. Claude Code starts reading files, Cursor’s agent starts planning, ChatGPT starts thinking. Now you have somewhere between five seconds and five minutes. Here’s how to spend them well.
Why AI tools keep you waiting
Coding agents do real work between your prompt and their answer: they read files, search the codebase, plan changes, edit several files and often run the tests. A one-line fix can come back almost instantly. A refactor across a few files can take a couple of minutes.
Ask a dozen times an hour and your day fills up with small gaps. Most of them are too short to start anything real, and too long to stare at a spinner.
Why the feed is a bad fit for a 30-second gap
The reflex is to open a feed. The trouble is that feeds are designed to have no natural stopping point. You open one “for a second”, and look up when it has decided you’re done, not when your agent is.
Then there’s the trip back. A feed fills your head with things that have nothing to do with the task, so you return to the diff having half-forgotten what you asked for and what you meant to check. That’s exactly when a wrong assumption slips through review.
A good break for this kind of gap is short, has a clear end, and leaves nothing behind to think about afterwards.
How Bitty fits the gap
- Start your AI as usual.
- Open Bitty: click the little buddy on your desktop, or press Option+Shift+B on a Mac (Alt+Shift+B on Windows and Linux).
- Pick the wait: a 5-second Blitz, 15, 30 or 60 seconds, or a 5-minute Boss Fight. The round fits inside it.
- Press Esc whenever you’re ready. Your round is saved and you’re back at your work.
Bitty doesn’t connect to your AI tool and doesn’t try to guess when it’s finished. You decide when to play and when to stop. That’s also why it works alongside anything: Claude Code, Cursor, ChatGPT, Codex, Copilot, a slow build or a long test run. If you like, start a rough wait timer from Bitty’s menu bar or tray icon and it’ll nudge you when the time is up.
What you can play
Six small arcade machines, each with puzzles generated fresh every round:
- Mental Math: quick sums, percentages and estimates
- Memory: digits, grid sequences and colours
- Pattern: sequences, symbol cycles and odd one out
- Logic: ordering, time zones and tiny syllogisms
- Code Prediction: what does this JavaScript or Python print?
- Reaction: wait for GO, then hit it
Each game keeps its own difficulty that follows how you play. See every game with real example puzzles, or play a 15-second round in your browser right now.
Nothing about your work leaves your screen
Bitty has no AI inside and doesn’t need one. It never reads your code, files, terminal, clipboard, browser history, screen or AI conversations. You don’t need an account, and the games work offline. Sign in with Google only if you want your progress synced or want Premium. The details are in the privacy policy.
Other good things to do while your AI works
A game isn’t always the best use of the gap. A few alternatives that also end cleanly:
- Stand up and stretch. Roll your shoulders, look at something far away for a bit. Your back and eyes will thank you.
- Review the last diff. If the agent already changed something, read it now while the context is fresh.
- Write the next prompt. Draft the follow-up, or note what you’ll check when the answer lands: tests, edge cases, naming.
- Get some water. A long agent run is a perfect excuse.
- Tidy one small thing. Close stale tabs, clear a notification, rename a branch. Small and finished.
Bitty is for the gaps where none of those fit, or when you just want a bit of fun.
Try it free
Bitty is a free download for macOS, Windows and Linux. The free plan includes 5 challenges a day plus unlimited 5-second Blitz rounds, every core game and three buddies. Premium adds unlimited play, Boss Fights and more.