Anthropic had the best AI product on the market. Claude Code with Fable was, without question, the best thing that had happened to a power user in years. And what they have done with that lead is a textbook case of how to destroy your most loyal users' trust through cutbacks, patches, and promises with an expiration date. Two limit extensions in two weeks. Each announced as "exceptional." Each ending with the same disclaimer: "this won't last." Generosity? No. Panic.
TL;DR: The No-Fluff Summary
- Two extensions in two weeks: Fable moved from July 7 to July 12, then from July 12 to July 19. The pattern is now predictable, every extension comes with the same warning: "but this will end."
- Prohibitive pricing after July 19: Fable only via API at $50/M output tokens, double what Opus 4.8 costs. Frontier AI for the financially privileged.
- Sol as a real alternative: OpenAI launched GPT-5.6 Sol included in the ChatGPT subscription. Advanced users are already migrating their workflows.
- Active class action lawsuit: users sued Anthropic for bait-and-switch practices on usage limits.
Two Extensions in Two Weeks: What Just Happened With Fable
Anthropic reintroduced Claude Fable 5 globally on July 1, 2026, after a 19-day suspension triggered by U.S. government export controls. But it didn't come back the same.
It came back capped at 50% of weekly limits across Pro, Max, Team, and Enterprise plans. With an expiration date attached.
Original expiration: July 7. Which became July 12. Which became July 19. Two extensions in fifteen days.
And in every email, every announcement, the same disclaimer. Anthropic is always careful to clarify, yes, for now, but this will end eventually. Which is frankly absurd.
The timing? OpenAI launched GPT-5.6 Sol in the first week of July, a frontier model included in the ChatGPT subscription. Sol is landing well, performing well. I'm using it myself. I'm genuinely considering closing my Anthropic account entirely.
Anthropic justifies the extensions due to "considerable user pressure." Translation: people are furious and they don't know what to do about it.
The Squeeze Play: A Timeline of Anthropic's Cutbacks
What's happening has a pattern. Line it up sequentially and the picture is damning.
Claude Code with Opus dominates for months without a serious rival. Anthropic has a product its users love. Workflows get built. The ecosystem takes root. The community grows. Everything looks great.
On May 6, 2026, Anthropic doubled the 5-hour Claude Code limits for all plans. It looks like a gift.
On May 13, they raised limits an additional 50%. With an expiration date. Now extended to July 19. We saw this coming when they tested cutting Claude Code Pro: the pattern is give first, then negotiate the take-back.
What the community had been sensing was that Anthropic was going to charge separately for the Agents SDK. That got derailed because the U.S. government banned Fable for export controls on June 12. Double trouble: you can't hit users with two cutbacks at once without torching your entire base.
Fable comes back July 1, but neutered: 50% limits with an expiration. The post-July 19 plan: API-only at $10/M input tokens and $50/M output tokens. Double what Opus 4.8 costs.
Anthropic has spent months squeezing its users, pulling Fable, bringing it back at punishing API prices, gutting Claude Code limits, and shrinking every margin they had.
The community pushed back. On June 14, a class action lawsuit was filed, alleging Anthropic misled subscribers about usage limits in a bait-and-switch scheme. Updated July 2, 2026.
Anthropic's plans clearly haven't changed, they're waiting for the right moment to push through what they've always intended.
The Compute Deficit: Why Anthropic Cuts (and Why It Doesn't Justify Anything)
Anthropic isn't cutting out of malice. It's cutting because it has no choice. The problem is structural: they don't have enough compute capacity to sustain Fable inside a subscription.
OpenAI locked in massive compute when prices were favorable, Anthropic didn't, and now they have to charge for it. OpenAI has also diversified with an advertising model that lets them subsidize frontier models at the subscription level. Anthropic runs on subscriptions and API revenue alone. Slotting a $50/M output model into a $20/month plan is mathematically unsustainable at scale.
The case for Anthropic? That they're doing the responsible thing: not subsidizing a loss-making product indefinitely.
Fair enough. But then don't launch the model inside a subscription, hook people for months, build an entire ecosystem on that promise, and then strip it away piece by piece. The unforgivable part is the pattern: hook them, build dependency, then cut.
And I'd bet this is the key point that Anthropic's defenders keep missing: the company made the strategic mistake of not building real switching costs. Their Claude Code skills and configurations are portable. They didn't build a closed ecosystem because their founding story was built around being the open, ethical AI company. That same openness is now what makes migrating your entire workflow to another model a matter of days, not months. They built an excellent product without building a moat. And when they started charging as if they had one, they discovered it didn't exist.
Anthropic is going from the brand everyone wanted to the one everyone's walking away from.
And yes, someone will say: "the extensions prove they're listening to the community." No. Listening would mean changing the plan. This is postponing the execution of the same plan until it's convenient. Every communication makes it clear the outcome will be the same. Condescension with a countdown, that's what this is.
They're just delaying the reckoning that's coming.
The Irony: Your Own Gift Is Their Exit Ramp
The Fable extension to July 19 isn't retaining anyone. It's giving users time to migrate. And the ones making the most of that time are exactly the users Anthropic can least afford to lose.
Power users are already using Fable to export their skills and entire ecosystems over to Sol.
It's like giving a week's notice to an employee with server access. What do you expect them to do? Exactly: pack their things and walk.
They seem to believe these extensions will keep users satisfied enough not to migrate.
And yes, "Sol isn't exactly Fable, the comparison isn't fair." There may be technical differences. OpenAI may be subsidizing Sol at a loss to gain market share. But for the professional user who's migrating, functional equivalence is enough. And if OpenAI is subsidizing, at least the user benefits while it lasts. Anthropic neither subsidizes nor offers an alternative.
If you've built your workflows with Claude Code, you already know that depending on a single AI provider is a genuine business risk. We covered it in depth in our guide to automating content with AI at scale: diversifying is not optional.
What To Do Before July 19
If you're an Anthropic user, the clock is running. No "wait and see." What's coming, they're already telling you in every extension. Act on data, not frustration.
- Export before July 19: skills, Claude Code configurations, workflows. Everything portable, move it now while Fable is still running in production.
- Evaluate Sol in parallel: don't wait for Fable to expire before testing. Run your real use cases and measure. Migrating with a safety net is far less traumatic than a forced migration.
- Don't cancel out of anger, cancel on the numbers: if Sol covers 80% or more of what you were doing with Fable in Claude Code, the math doesn't lie. If not, wait, but have Plan B ready.
- Always diversify: maintain at least two functional workflows across different providers. Single-provider AI dependency is a real business risk, not a consultant's talking point.
Frontier AI for the Financially Privileged
Put the numbers side by side. Sol, a frontier model, included in the ChatGPT subscription. Fable, a frontier model, at $50 per million output tokens via API-only after July 19.
The numbers speak for themselves: frontier AI for those who can afford it. And that directly contradicts everything in Anthropic's founding narrative about safe and accessible AI.
In the end, this is AI for the privileged, not for everyone.
And you know what's most ironic? The patches aren't stopping anything. Each extension gifts another week to the people who know best how to use your product, so they can take it somewhere else. And those people don't come back.
Anthropic can keep stretching it. From July 19 to July 26, from July 26 to August 2. It doesn't matter. The plan hasn't changed. They're just delaying the reckoning that's coming.
