The model is only part of the story.

AI can answer a question with remarkable fluency. I’m interested in a question that comes before the answer: what does it actually know about the situation?

In healthcare, a conversation is part of the picture. A doctor also works with history, observations, reports, experience, and a sense of what has changed. Some of that information is in a record. Some is on paper. Some is visible only in the room.

Conviction in AI pessimism and limits

“If a human with every resource at their disposal can’t do it, AI can’t either.” That is my skeptical starting point.

I do not see AI as a way around missing evidence or the limits of what can be known. If the information needed for a decision is missing, an AI system should make that gap visible. Confidence should not disguise missing evidence.

I also want to stay open to AI helping people find patterns they would otherwise miss. The boundary of what is possible is something to explore through evidence.

The data layer is the starting point

Could we bring the relevant context into a layer a clinician can inspect and control?

That could mean conversations, earlier prescriptions, report images, visual observations, or operational signals. It could involve a phone, a wearable, smart glasses, edge/IoT, or software a hospital already uses. I’m not committed to a device. I’m interested in what information is useful and how to capture it responsibly.

Voice tools such as Wispr Flow are an interesting reference for natural interaction. Meta-style glasses are a possible window into a clinician’s point of view. Neither reference means a partnership or an integration exists.

What would make the context useful?

I want to ask three questions:

• Can a doctor see where the information came from?
• Can they correct it, question it, and identify what is missing?
• Does the resulting intelligence help in a real clinical workflow?

Those questions connect all three directions for Convertase: clinics that remember, doctors working with richer context, and hospitals that can understand the work around care.

The first product remains open. The data layer is the thread I want to follow.