The Model Continuity Checklist

Last month a US export order switched off two frontier models worldwide for almost three weeks. The companies that stayed running were not the ones that bet on the right model. They were the ones that never had to. Here is how to become one of them. Download the one-pager above and print it, or check the boxes right here next to your architecture diagram.

1. Map your exposure

List every model your business runs on, one line each. Next to each one, write the name of the alternative you have actually tested carrying that load, not the one you assume would work.

Model you run on	Tested alternative	Days to switch

Every blank in the middle column is exposure, drawn to scale. Now you can see it.

2. Run the four moves

01 Put a routing layer in front.

Models are called through one routing layer, not hard wired in a hundred places.
Switching a model is a configuration change, not a rebuild.

02 Keep a real second model, tested in anger.

You run a second model from a different vendor.
and from a different country.
It has been run against real traffic, not a demo.

03 Hold one open weight model you can run yourself.

At least one open weight model runs on hardware you control.
It was stood up and tested before you needed it, not during the week you lose your primary.

04 Put portability in the contract.

You have the written right to export your data on demand, in a usable format.
and everything you have built on top of it.

3. Score the one that matters

Pick the model box you would least want to lose. If it went dark tomorrow, still working perfectly, because a government or a vendor decided you could no longer use it, how many days until you are running on the alternative?

If there is no number, because there is no alternative, you have just found this week's work.

I should disclose an interest. The company I helped start, Swa, works on model portability. That is my bias, on the table. Every one of these four moves stands whether you ever look at what we make or not.