TypeSafe employee here,
This could change the narrative around AI and safety. The Jev model isn't a reasoning model, it doesn't have hidden thinking tokens. This is how to bring engineering back into the conversation, design your reasoning systems as code, that way they are explainable. Explicit reasoning is the interface for humans to constrain and control, this opens completely different approaches to alignment.