🏚️ What a crazy email to wake up to.
If you haven't been following the news: Fable was out for 72 hours before a letter from the US government forced Anthropic to pull the most powerful model it had ever shipped. Talk about a press release. 😬
If you spent this whole week wiring Fable into a new workflow, recording a tutorial on it, or building a side project that leaned on it, first off, sorry to hear that. 🥲
But hopefully this helps all of us realize something: frontier models are not stable primitives. Unlike past versions of software, these models aren't stable releases you can safely build on top of the minute they drop. A model can vanish over a security threat, a price change, a policy shift, or a single letter from the Commerce Department.
So the things you really want to build shouldn't be tied to one specific model or one company. They should be adaptable. The thing worth investing in is the muscle underneath it all: the judgment, the process, your ability to get close to the same result out of any model. And all of that lives in your workflows, not in the model.
As much of a Claude fanboy as I am, this is a great lesson. Whatever you build on top of a single frontier model, you're building on rented land. 🏚️
So don't fall in love with specific models. Work on the skills that let you stay model-agnostic, because those are the ones that survive the next letter from the government.
Related: Building that model-agnostic muscle is exactly what I help engineers and teams do. If that's useful to you, it's all here: alextong.me/services →
