Some reactions to the Mythos and Fable restrictions jump straight to "we need sovereignty," then to "which platform provides it." Both come down to a set of decisions you weigh, based on what matters to you and the outcome you want.
Two decisions I would be weighing right now if I were responsible for an enterprise stack.
First, does anything you run survive a model going dark? I wrote a while back about routing queries across a fleet of models to control cost. That same routing layer can be used for continuity. A cheaper tier was a budget decision. A tier you host yourself can be the difference between more planning upfront and a stopped workflow.
Second, where does your accumulated learning live? If the judgment your teams have poured into a system is baked into a model you do not control, you do not own that judgment. You are renting it from whoever can switch it off.
Companies are announcing what I would describe as an enterprise harness. How it gets built decides where that line falls. One example. You move enterprise memory and tuning to a cloud platform, then apply it to an open weight model you host yourself. You control the model. The structure around it stays platform specific.
Does your architecture buy resilience or dependence?
Written by Duane Grey
AI Strategy & Implementation
Independent AI consultant helping companies cut through hype and deploy systems that produce real results.