On June 12, Anthropic — the company behind Claude — got a directive from the U.S. government and had to shut off two of its most powerful AI models, Fable 5 and Mythos 5, for every customer, overnight. A security and export-control concern. The AI world lost its mind. (Anthropic said it disagrees and is working to restore access — and that every other model kept running. Here’s their statement.)

If you felt a flicker of “wait — is the AI thing I’m leaning on going to vanish too?” — good. That instinct is worth two minutes.

A model getting yanked overnight is a five-alarm fire if your business is hitched to that one model. It’s a shrug if it isn’t. And the same announcement that pulled Fable said the quiet part out loud: every other model kept working.

The operators who shrugged are the ones who did the boring foundational work — they wrote their business down into the context files we keep talking about. To them, a model is a tool brand, not a dependency. The fancy one disappears? Point the same context at a different model and keep moving — same as grabbing a different impact wrench when one dies mid-job. The ones sweating are the ones who bet everything on having the newest, biggest model, and now they’re refreshing a status page instead of working.

Here’s the trap, and it’s an easy one to fall into. The AI companies sell on benchmarks — “ours scores two points higher on this test” — and the press covers each release like a heavyweight title fight. So you start to believe the model is the product, and that picking the “wrong” one means you’re falling behind.

You’re not. For the actual work you’d hand an AI — drafting a bid follow-up, cleaning a spreadsheet, summarizing a contract, writing a safety brief — every serious model on the market is already more than good enough. That two-point benchmark gap is invisible in real use. What’s not invisible is whether the AI knows your business. A “lesser” model loaded with your context beats the world-champion model that’s never heard of you. Every time.

Stop chasing the model leaderboard. It’s the most over-covered, least useful story in AI. Which model is “best” this month changes constantly — and as Fable just proved, the biggest one can be gone by Friday. What doesn’t change is your context: the written-down version of how your business runs. Build that, and the model becomes a swappable part.

Picture it concretely. Say the model you use every day went dark tomorrow morning. If your standards, your job types, your templates, and your rules live in a folder, your move is thirty seconds: open a different AI, point it at the same folder, keep going. If all of that lived only in your head and in a chat history with one tool, you’re starting over. Same event, two completely different mornings — and the only difference is whether you did the context work ahead of time.

So next time a breathless AI headline crosses your feed, run it through one question: does this change what I can do with my own context? For “they pulled a top model,” the answer is no. For most headlines, the answer is no. That’s the good news — you get to ignore about 95% of the noise and keep working.

Once your context exists, the model is the least interesting piece of the whole setup. The interesting part is what you stack on top — the thing that drafts the bid, pulls the job costs, has the Monday brief ready before you sit down. Those workflows don’t care which engine is underneath; swap the engine and they keep running. Models come and go. What you build on your foundation compounds.

That’s the whole game, and it’s why the headlines can’t rattle you: you’re not betting on a model, you’re building an asset that outlives any of them.

Don’t marry a model. Build the context, stay swappable, and let the leaderboard drama be someone else’s Tuesday.

Reply and tell me: what AI headline made you wonder if the ground was moving under you? I’ll translate it — operator to operator.

Next week, back to a hands-on win you can use the same afternoon.