AI is scaling faster than its control layer
IBM has named the enterprise AI control gap. This piece argues that the missing answer is a behavioral integrity layer for the AI system in operation.
thinking
essays, field notes,
and public thinking
Toriel's canonical thinking lives here first. This library gathers the essays, field notes and signals that make the architecture legible: control, continuity, identity, governance, and what intelligence will need to become.
featured essays
These three pieces introduce the foundations of Toriel's work: 53 on control, 41J on continuity across change, and 47 on identity beyond memory.
IBM has named the enterprise AI control gap. This piece argues that the missing answer is a behavioral integrity layer for the AI system in operation.
Portability and replaceable models are becoming enterprise requirements. That changes where continuity has to be governed.
Retained facts, familiar tone, and recalled preferences do not by themselves amount to a system remaining itself across change.
latest thinking
New essays and field notes will be added here as the category develops — from behavioral integrity and AI governance to model interchangeability, continuity infrastructure and signals from the market.
IBM has named the enterprise AI control gap. This piece argues that the missing answer is a behavioral integrity layer for the AI system in operation.
Portability and replaceable models are becoming enterprise requirements. That changes where continuity has to be governed.
Retained facts, familiar tone, and recalled preferences do not by themselves amount to a system remaining itself across change.
As companionship and emotional support become mainstream AI use cases, continuity and verification stop being optional architecture.
The most important AI changes are not always the ones announced in release notes.
A leaked “soul doc” is less interesting as a curiosity than as evidence of a deeper architectural truth: AI behavior is shaped by layers most users never see.
Why prompt routing, wrapper design and orchestration define the AI experience, not just the model.
If AI becomes the next operating layer of human life, the central question is not only how intelligent it is. It is who the relationship belongs to.