Fable Is Gone (Wait, No!)

Fable is gone.

Well shit.

My Whatsapp Group has been abuzz with working out what the hell to use our final moments with Fable on. And then mourning.

Or so I wrote yesterday before we got the news of a reprieve. A stay of execution:

This obviously messed up the newsletter a bit. BUT everything below is still accurate. It’s just….delayed by 5 days ha!

As of Tuesday July 12, 2026, Anthropic is moving Claude Fable 5 out of normal paid plans and into usage credits. The API page has it at $10 per million input tokens and $50 per million output tokens.

Now, you might be wondering if that’s a lot. Or even what is a token? How many input and output tokens could a mere mortal use in a day? Let’s run the numbers quickly.

Here for example is me using 900M tokens on Codex in a day (and what a day ‘twas!):

That ONE day, if I had been used that many tokens on Fable would cost me between $9,000 and $45,000

Coding tasks tend towards the input so probably around $12,000-15,000 for the day.

I think my business partner Harms would have been a bit miffed.

Using Fable from now on will be the reserve of companies or the very wealthy.

And I get why people are annoyed. For the last week, a lot of people finally saw what these models can do when they are given actual work.

I used it on finance stuff I had been dodging. I used it on a book project I did not want to outsource to slop. I used it on business systems that had been stuck for months. And Fable just tore through everything.

And now the clever thing is behind a meter. A very expensive meter. Bugger.

Zoom out

The pricing story will get all the shouting and screaming Fair enough. People got a taste of Fable, then Anthropic put it behind usage credits. Cue moaning, mourning, and screenshots of token maths.

But we need to step back and remember the last few years.

A frontier model arrives. Everyone falls in love with it for about five minutes.

Then the next lab catches up, the price drops, the open models trail behind, and what was once science-fiction becomes boring infrastructure.

We get VERY used to our models very quickly.

We will have other models like Fable shortly. Probably this week with ChatGPT Sol. Or if not definitely in the coming months.

The direction is obvious. Fable feels rare and special because we are standing too close to it. In a year, probably less, this level of intelligence will feel normal.

Think about this for a moment:

One day, in the next 2-3 years, we will have a Fable level model running locally on our iPhones.

Read that again and really let it sink in.

Our models are getting better, cheaper and smaller all of the time. ChatGPT 3.5 is just 3.5 years old. And it was TERRIBLE (seriously, go try it on the Playground for a laugh).

It’s taken 3.5 years to come from ChatGPT 3.5 to Fable. What will we have in 3.5 years from now? THAT is what we need to keep our eyes on here.

What happens when its cheap?

When a model is expensive, you ration it.

You use it for the big judgement calls. The ugly personal problem. The business mess where being 20% smarter changes the answer. That was the point of last week's Fable question issue: do not waste the expensive brain on tiny tasks.

But when the price falls, behaviour changes. What happens then?

We will one day have this Fable near-human level intelligence for basically free. On our phones. On our local devices.

That brings us to the darker side.

If you have used Fable over the last week.

And if you can imagine Fable being made free and unlimited.

Now tell me that jobs are safe.

It becomes a very difficult argument to make.

And remember businesses do not even need a model to be free. They need it to be cheaper than the human task it replaces. And when Fable-ish capability gets fast, cheap, and boring like a utility, a lot of white-collar work starts looking wobbly.

The AI doesn’t have to take entire jobs. No need. Remember that jobs are bundles of tasks. If AI takes 30%, 40%, 50% of those tasks, the job title might survive while the income, headcount, and employee bargaining power get quietly chewed up underneath it.

This is where I get annoyed at the "it'll all be fine" crowd. They accuse people like me of being a Doomer. No - I’m a realist. And after seeing what Fable can do I’m doubling down on that. Our comfy 9-5s are not safe.

Build something you control

You cannot rely on your company to handle this well. If you are employed you need options.

Maybe your employer has a clear AI strategy. Maybe they have tooling, training, governance, a proper plan, the lot.

Maybe.

Or maybe they faff around for two years, buy a licence, form a committee, and then panic-cut costs when a competitor uses AI faster than they do.

If you work inside that machine, you do not control the machine. Sorry.

Or shit maybe they do know what they are doing and you still aren’t safe. Because the first moment they can drop you and replace you with AI they will. Corporations want our loyalty but the honour does not extend in both directions. They will train the AI on your daily work (screen recordings, emails, meeting notes, any and everything they can get their hands on) and then switch you out for an agent that does the task for 100x lower cost.

And maybe even better than you…

So what do we do?

So build something you do control. A side income stream. A consulting offer. A tiny first product. A niche audience. A service. Something you build yourself and control.

I’ve actually built a whole 10 week course on building a business with AI. Totally free, no need to even register.

Is it easy? Nah, not easy. But increasingly we don’t have much of a choice.

We either wait for the sword for fall. And then complain that we didn’t see it coming (we did).

Or we take responsibility and get on with building something that we control.

If you are looking for a starting point I say lean into the chaos! Let the confusion and fear of AI be what drives your business. Instead of being driven by it. This is why I keep banging on about workshops and advisory work. EVERY business on earth right now knows they need help with AI.

And they have no idea what to do about it.

You do. You are in a very very small % of people who are paying attention to AI. I know this because you are here reading a debrief about Fable. That’s pretty niche info you know.

If you can explain this stuff clearly and turn chaos into action, that is valuable to businesses. Extremely valuable. Several thousands pounds/dollars an hour valuable.

I’ve got 250+ students are building around that model now. Workshops, consulting, implementation, advisory. If you want to know more I’m doing a live webinar tonight at 6PM London time: aiwithkyle.com/webinar.

To the Task,

Kyle