Focus Space v2: What Happened When I Gave the AI a Free Hand

I built the first version of Focus Space step by step: a spec (PRD) first, then the implementation stages one after another, properly by the book. While I was working on it, Anthropic, the company behind Claude Code, released a new model, Claude Fable 5. There was a lot of buzz around it, with plenty of people claiming it could build an entire app from scratch off a single prompt.

I decided to test that, so I handed the model one big prompt with a task: write Focus Space from scratch. With one caveat: this time I deliberately widened the scope. I didn't copy the assumptions from the first version, which had been intentionally trimmed down to the absolute minimum. Instead I described my fuller vision to the model: I wanted the focus minutes not to simply tick away, but to add up to a journey. I wanted the spaceship to actually be going somewhere, discovering new star systems, and the whole thing to feel more like a game than a productivity app.

What came out of it

I was pleasantly surprised by the result. Claude Fable 5 produced a completely different interface, simpler in execution. It's not my ultimate dream look yet, but it's an interesting concept.

The model added a map of star systems: every minute of focus is light years the ship travels, discovering star after star. On top of that, it used real stars near Earth and laid the route out in the shape of an unfolding spiral. Then there's a currency of its own, Flux, which you can spend on upgrades (changing the interface colour, for instance), plus a rank system, military grades that go up along with the number of light years flown. After each hyperspace jump you get a funny write-up about the systems you've discovered, jokes included. The whole thing comes off more like a game than a dull Pomodoro timer. That's a step in the right direction.

I didn't review the code itself, it could probably use a refactor, but that's a separate matter.

You can check this focus-game version out here: Focus Space Game

What I learned

There are completely different ways to approach the same idea. This second iteration has left me wanting a third version, maybe even a fork, something different again.

I'd like the eventual app to have a stronger game element: gamification, a currency, some kind of stake attached to the minutes of focus, much like Forest, where a forest grows and points can be traded for decorations. I also think it's missing characters, something along the lines of the astronaut mascots in Kerbal Space Program. Having them around would work as a gentle nudge: don't switch to another app, or you'll be doing your crew harm.

I don't know when yet, but I want to come back to both of these iterations and build a third version, as engaging and as interesting as I can make it.