The day Fable came out, I sat down to work and thought something was broken.
A little popup told me Fable 5 was available to use — but I couldn't find it anywhere. A couple of hours later it was fixed, and I thought... okay, what's something big I can test this on?
Fable's supposed to be really good at one-shotting things: thinking through each step, getting the right agents doing the right things, and then testing it all itself. So I figured — let's try this out. Let's see how well it can make a game I've wanted to build for a while.
I sat down and started a prompt. I included screenshots from my previous attempts, sketches of flows, outlines from parts of my design doc — I gave all of it to Claude, turned on bypass permissions, pointed it at a folder with some CLAUDE.md files, and left for a bit.
It Worked Out of the Box
When I got back, I was greeted with instructions for spinning up a local server for the game. I opened it in my browser and... let me tell you, this thing worked extremely well out of the box.
Not perfect, mind you. But everything worked to some extent, and it followed my game design docs, my ideas, even the screenshots of potential UI elements. It built out some of the gameplay and even added a couple of short tutorial notes to help players get started.
To put that in context: I've spent hours and hours over the past few years trying to learn Unity, Godot, and other engines to make this idea work — and I always got stumped on dumb functionality. Things like getting two cards to stack when you drag one on top of the other, or building a system to detect which cards are in a specific pile.
Fable didn't just get all of that right. It added the little UX touches too — picking cards up, dragging them, right-clicking to split a stack, snapping cards to each other. Heck, it even added sound effects.
The Part I Always Love
And then it was time for the part I always love: paying attention to the small details.
Yeah, yeah — it's way too early in the process to nitpick. That's what part of me was thinking. But at the same time, I was able to polish those details while also pushing the concept of the game forward.
In design, there are two kinds of people: those who can look at a wireframe and understand what the thing is meant to be, and those who need the final version in front of them to get it. I've worked with plenty of clients in both camps. And sitting there, I realized — it's actually kind of nice to not be stuck in wireframe-and-flowchart mode for once.
The game wasn't “final,” but it was a lot further along than wireframes. Combine that with real gameplay logic, and I had a working, semi-polished prototype I could actually play to gauge the one thing that matters most: is it fun? And believe me — so many things I was sure would be fun turned out to be boring or annoying once I actually played them.
Which is where I had to take a step back and make a plan.
Rushing ahead is a blast. Claude makes it so easy to design, prompt, build, test, iterate, and repeat — it even took on the design system I'd built and helped me dial in the game board settings like camera angle and zoom levels. But now I had a problem.
So far, the game was built on a backend and rule set that lived mostly in JavaScript and JSON files. As a designer, those two things are about as far from a useful “game design file” as you can get — lists and code that run the thing and hold the rules. Turning this into the game I actually wanted to build — ideally with some kind of interface for crafting levels, recipes, upgrades, and progression — was a whole other thing.
And that's the hidden part I've been circling, and the real point of this post.
It's Easy to Have AI Build Things That Look Finished
Fable 5 is amazing, but it rolls a lot of steps into one. Things I'd normally have to do myself in Opus — setting up each of the systems, building tools to work with them (even if it's just spreadsheets for recipes and strings), building out the map and individual elements — Fable just blew past all of it and produced something that nailed what I was hoping for. But in doing so, it also created a big chunk of debt work that I'm now dealing with: essentially building out the “engine” and the “studio” parts of this myself, and getting those working.
Honestly, that's not much of a drawback, and I'm happy to do it. Building a custom studio to edit the game experience, and working on the engine side to render and optimize it, are both fun. And going forward, knowing which steps Fable will skip — the ones I'll have to rebuild — will absolutely reshape how I prompt.
But for quick proofs of concept? Fable 5 has shown it can validate an idea fast and help carry it from the “wireframe” version into something other people can look at and have that same ah-ha moment — even if they don't speak wireframe.
And being able to see progress, prompt a change, and watch it show up right away has done something I didn't expect: it's made me want to work on this more, and get it out for others to test and shape. Which is really what I'm after.
Designing in the Dev Environment
Because while I love the design process — love designing, love building — at heart I'm a product designer. I like building products: design, development, testing, and feedback, together. I just happen to have the most experience in design and UX, which flavors everything. But I'll happily sacrifice my perfectly organized, perfectly documented Figma files in the name of actually shipping something into the world.
The more I use Claude to build the things I used to only design, the less I rely on getting everything perfect in Figma. I work more in the development environment now — taking the actual code, seeing what it does, and influencing the design there instead of in Figma.
I still love Figma, to be clear. It's still my preferred tool. I just find I only go there when I need to solve something specific, when I need to show Claude what I'm talking about, or when I need to design real assets for the app — icons, illustrations, that kind of thing.
Build Your Thing
So here's my pitch: take the idea in your head, write it down in as much detail as you can, and work with Claude or Codex to break it into chunks and build it. I've learned more about actual iOS, web, and game development in the past six months of building with Claude than I did in years of grinding through tutorials and courses — all of which show you how to build someone else's thing, their way.
Building your thing is like crack. It's addicting.
If you haven't tried it... please do.