Claude Fable 5: the most capable model you can touch, and what to build with it
The most capable AI model the public can use. The video covers what it is. This page covers what people built with it and what you should point it at.
Claude Fable 5 is the most capable AI model the public can actually use, Anthropic's frontier Mythos brain with safety guardrails bolted on so anyone can touch it. Give it a job and it plans the work, does it, and checks itself for hours without you watching. I break down what it is in the video. This page is about what to point it at.
What people already built with it
In the first 72 hours, before it got pulled, feeds filled with the same reaction over and over: I can't believe it just did that. Some of the real ones:
- Three playable games from three one-line prompts, built by Wharton professor Ethan Mollick in Claude Code: a Snake clone, a lantern-lit tunnel maze, and a game built on Rilke's poetry.
- A full explorable Hogwarts in one shot. Classrooms, the Great Hall, the Quidditch pitch, all of it.
- A navigable Yosemite built to scale from real NASA elevation data, with roughly 266,000 procedural trees and custom water shaders placed on every waterfall.
- A working V8 engine CAD model in under ten minutes.
- A native iOS calorie tracker from a single prompt, an actual app and not a web page.
- A Minecraft clone with multiple biomes, a day-night cycle, ores, and caves, in about 20 minutes from one prompt.
- A cross-platform analytics dashboard that pulls its own numbers out of sites with no public API, by watching the hidden requests the site makes in the background and copying them.
What you should build with it
The pattern is the same every time: someone described a thing and then the thing existed. If you make content, run something, or just have a list of tools you always wished existed, this is where to point it.
If you make content
- A personal analytics dashboard that pulls every platform onto one screen, so you stop opening five apps to see how a post did.
- An agent that takes one video or article and reformats it for every platform on its own while you sleep.
- Interactive toys as content. A playable mini-game tied to your topic gets shared far more than another static post, and people built full games from a single sentence.
- A second brain over everything you've ever made, so you can ask what your best hooks had in common and get a real answer instead of a guess.
If you run a business or a brand
- A real landing page at designer quality, from a description or even a screenshot of a page you already like.
- The internal tools you keep putting off: a lightweight CRM, a booking flow, a dashboard your team will actually open.
- The big migration or cleanup nobody wants to touch. This is the Stripe use case, scaled down to your mess.
- A custom app for one specific job, instead of paying monthly for bloated software that does that one thing plus fifty you never use.
Just for you
- A wall-display dashboard for your kitchen: calendar, weather, whatever you want on the screen. Someone built one for a 77-inch OLED.
- An app to run your own house: your music, your lights, your routines, in one place.
- The weird passion project you'd never pay a developer for. An explorable universe, a game only you and your friends will play, a tool that kills one tiny daily annoyance.
How to actually use it
- 1.Use it inside an agent, not a chat box. Claude Code, or a similar harness, is where Fable earns its price. That's what lets it run for hours instead of one message.
- 2.Describe the outcome, not the steps. Its goal mode lets you state an end result and let it iterate until it gets there. You don't need to know how it's built.
- 3.Turn the effort up for the hard stuff. Higher effort means it thinks longer and pushes further before it stops.
- 4.Run jobs in parallel. People handed it nine separate workstreams in one night. If you have a backlog, line up the prompts and let them run at once.
- 5.Keep a cheaper model open for the small stuff and save Fable for the jobs that earn it.
The bottleneck now
The gap between people who can build things and people who can only picture them just got a lot smaller. Having an idea and shipping it used to be two different skills, and Fable erases most of the distance between them. So the bottleneck is no longer the code, it's whether you have a list of things worth building. If you don't, start the list.
