The World’s Most Powerful AI Is Back, and You Have Until July 12 to Use It at No Extra Cost
The government shut it down three days after launch. Now it’s restored on every paid Claude plan, the trial clock got reset, and here are the three things I’d build before the window closes, starting
Anthropic just moved the deadline. Fable 5’s included window was set to close yesterday, July 7, and today the company announced on its official X account that paid subscribers get five more days, through July 12. That’s on top of the comeback itself: the Commerce Department lifted the export controls on June 30, Anthropic flipped the switch on July 1, and the most capable AI model the public can touch is back in the Claude menu like the shutdown never happened.
I’ve covered this story from the start. The shutdown three days after launch, how export controls actually work, and what a trillion-dollar company getting switched off by one letter should teach anyone building a business on AI tools. If you missed those, they’re here and here. So I’ll keep the recap short and get to the part that matters, which is what I’d actually DO with this model in the next five days.
How it came back
Anthropic spent the blackout proving the government’s case didn’t hold. Its own testing showed that older Claude models, GPT-5.5, and even Kimi could find the same software vulnerabilities that triggered the ban. The exploit that got it pulled off the market was something half the industry could already do.
What satisfied Commerce was a new safety filter, built with the government and Amazon, that blocks the reported technique in more than 99% of cases. The trade-off is a touchier model. Some legitimate coding and debugging requests now get flagged and quietly handed to Opus 4.8 instead, but Anthropic says it’s working to reduce those false alarms. Mythos 5, the version without the guardrails, stays locked to roughly 100 vetted US organizations.
Back in June I told you to watch HOW the government backed off, because if the models returned without anyone admitting error, the only thing established would be that Washington can reach in and flip the switch whenever it wants. The Commerce Secretary posted that his department “worked closely with Anthropic to analyze and approve Fable 5.” Cooperative tone, zero acknowledgment that the original case fell apart under testing.
Why that matters to you: a shutdown that ends with no admission of error is a shutdown that can happen again on the same thin evidence. The precedent is now set. One letter, no hearing, no proof required, and the tool your business runs on can go dark for three weeks or be removed entirely. Nobody involved paid a price for getting it wrong, which means nothing discourages the next one. That’s the operating reality you’re building in, and it should shape what you use this model for.
What you have until July 12
Through July 12, Fable 5 is included on Pro, Max, Team, and select Enterprise plans for up to half your weekly usage. This is the window’s second reset. The original one, June 9 through June 22, got swallowed by the shutdown three days in, and the July 7 deadline just got the five-day extension.
After July 12 it moves to usage credits at $10 per million input tokens and $50 per million output tokens, which makes it double the price of Opus 4.8 and the most expensive model Anthropic sells. So July 12 doesn’t end your access, just the ability to use it inside your existing subscription. Five days of included time on a model this capable is worth attacking with a plan instead of messing around with it for twenty minutes and wandering off.
What I’d build, knowing what I know
A few weeks ago I told you not to stake your income on the frontier model, and then a government blackout made the argument for me. Nothing about the model coming back changes that. It already disappeared once, and the mechanism that took it out is still sitting there, fully intact.
Which is exactly why the smart play this week is one-time assets. Projects with a beginning and an end that leave you holding something when they’re done. A finished product. A full audit of your business. If a letter takes Fable out again next Tuesday, everything you built during the window still belongs to you.
And this happens to be the work Fable was built for. Anthropic’s own framing is that the longer and more complex the task, the larger Fable’s lead over every other model. You hand it a big project, it runs the whole thing, and you review finished work instead of supervising every step. I felt that the first day I used it. It turned out landing pages in a fraction of the time Opus takes, and when I gave it ten setup and structure documents from a client’s business and asked it to evaluate them, it went through all ten in ten minutes and told me what was solid, what needed work, and what needed a rebuild from the ground up. That kind of review used to take days.
So that’s the bar for this week: big projects, handed off whole, finished when they come back. I picked three you can run with your included usage, each one ending with an asset you keep, and I’d execute them in this order.
1. Build a digital product from what you already know
This is the highest-value move on the list because it produces the one asset that can pay you back: something to sell. If people ask you for help with anything, you have expertise worth packaging. A build like this used to keep me in the chair for days, prompting section by section and stitching the pieces together myself. Fable takes the whole project, interview to outline to finished draft, and hands it back for your review.
Copy this prompt into Fable and let it run the project:
I want to build a digital product based on my existing expertise, and I want you to run the whole project.
Start by interviewing me. Ask one question at a time about my skill or knowledge area, who asks me for help and what they ask about, what results I’ve gotten for myself or others, and what my audience struggles with most. Keep asking until you can name one specific painful problem I’m qualified to solve.
Then propose three product concepts that solve that problem. For each one, give me the format (guide, template pack, mini course, or workbook), the core promise written as “This helps you achieve ___ without ___,” a realistic price, and what would need to be true for a buyer to get results from it.
Once I pick one, build the complete product. Outline it first and show me. After I approve the outline, draft every section in full. Include real steps, checklists where they serve the reader, and a simple implementation plan at the end so buyers act on this instead of just reading it.
Rules: no hype, no income promises, no manufactured urgency. Write plainly. Where my personal experience should carry a point, mark it [YOUR STORY HERE] and tell me what kind of story fits, because you don’t write my stories for me. Structure the final draft for packaging: clear section headers, and a note wherever a checklist, worksheet, or visual belongs, so I can design those pieces in Canva. When the draft is done, give me a one-page list of what’s left for me: my edits, my stories, the design work, my pricing decision, and where to sell it.
One piece of that prompt matters more than the rest: it leaves your stories to you. The fastest way to wreck trust with an audience is publishing personal experience a machine invented for you. Fable drafts the structure and the teaching, and the life in it stays yours.
Fable can also handle the design and turn your approved draft into a finished, ready-to-sell PDF. That prompt is in the paid section below.
2. Audit your whole business in one sitting
That ten-document review I did with my client about a month ago is the proof here. What Fable did for those documents, it can do for everything holding up your business.
Gather the pieces. Sales page, welcome sequence, lead magnet, offer description, and about page. Whatever exists. Hand all of it to Fable in ONE conversation and ask for a three-way verdict on each piece: solid, needs work, or rebuild.
The reason this beats reviewing pieces one at a time is that Fable holds the whole system in view at once, so it catches the places where your assets contradict each other. Your sales page promises one outcome while your welcome sequence teaches toward a different one, and your lead magnet attracts a buyer who wants neither. A document-by-document review misses that every time, because the contradiction lives between the documents rather than inside any single one.
What you walk away with is a prioritized fix list. If you’re early and only have three assets, you find out what to fix before you build more on top of it. If you’re established, you get the kind of full-system review a consultant charges hundreds or even thousands of dollars for, done in an afternoon. The exact prompt I’d use is in the paid section below, along with everything else.
3. Feed it a year of your work and ask what’s working
Fable holds about a million tokens in one conversation. In plain terms, that’s roughly 700,000 words, so a full year of your articles, posts, and emails fits inside a single chat with room to spare.
No other model I use can do this well. Shorter-context models forget the beginning of your archive by the time they reach the end, which means their “analysis” is really an analysis of whatever they read last. Fable reads the whole body of work and finds the patterns across it: which topics pulled subscribers, where your voice ran strongest, and which pieces you were sure would land but didn’t.
Export your year into one document, add whatever stats you have, and hand it over. You’re looking for the pattern your own memory is too close to see. The packaging instructions and the full prompt are below the paywall, because doing this one wrong burns a lot of included usage for a shallow answer. Substack writers, there’s also a walkthrough down there for pulling your full archive out of Substack, plus a prompt that turns the analysis into your next 90 days of article ideas.
Everything below is for paid subscribers: the design prompt that turns your product draft into a finished, ready-to-sell PDF, the copy-paste prompts for the business audit and the content analysis (including a Substack export walkthrough and an article idea bank prompt), two more high-value moves to run with your remaining usage, the order I’d run everything in, and the honest math on whether Fable is worth paying for after July 12.
