Shock and Disappointment, Fable

On June 9, 2026, Fable 5 was released. But on June 12, the U.S. government issued an order banning its use by foreign nationals outside the United States. Since it's realistically hard to single out only foreigners, Anthropic chose to block Fable 5 and Mythos entirely. It was a shocking and disappointing move.

From the day Fable launched on June 9 until it was blocked, I used it every day for three days—leaning on it heavily to surface improvements for the project I'm building and to map out where it should go next. As a Max20 user, my weekly limit happened to reset on Wednesday, the very day it launched, so I could make the most of my tokens from release onward. The moment I wrapped up work on Friday, my weekly limit was at 100% and my 5-hour limit had a little left. When the weekly limit runs out, you can't spend any more tokens even if the 5-hour window still has room.

When Fable first came out, I learned I could use it until the 22nd, and I figured I'd push myself to wring every advantage out of it until then. At the same time, even though the promise ran to the 22nd, I had a feeling something might cut it short for one reason or another. Anthropic has been through countless outages running its servers, and for months I'd been hearing that Mythos could be a double-edged sword for security—so I expected all kinds of variables once it actually launched. That's why I tried to apply Fable to my development work as much as I could while I still had it.

But actually living through a full block just three days in stirred up a tangle of complicated thoughts. sonnet, opus, fable… many more new and better models will keep coming. Developing and advancing AI has been underway for a very long time, yet at this point it's only been about a year since AI really began to touch daily life and work in earnest. Is it already threatening enough for governments to step in? The thought sent a chill through me.

After three days with Fable, my honest impression was that a better-performing model clearly lifts the quality of the work. I've been developing with opus for about four months, and Fable solved bugs opus had never managed, and I felt it figured out a lot on its own—things I'd worked hard to cage opus into with all sorts of skills, harnesses, and hooks. Even when communicating through prompts, I felt it understood my intent far better. It's a real shame I couldn't do more, since its token usage is twice that of opus. Even so, in the module-writing project I'd been working on, it caught around 50 bugs opus had struggled with (about 10 of them critical), greatly improving the module's stability. I wanted to put it to work on so much more, but for now the chance seems gone.

Someday the day will come when I can use the Fable model again. But once you've tasted something good, going back to what came before is very hard. For someone doing agentic coding with AI, Fable is a tool I want so badly right now. How wonderful it would be to use Fable with unlimited tokens.

Since I don't work in the security domain, Fable alone—not Mythos—is enough for me. I hope the day I can use Fable again comes soon.