Two stories this week, arriving from opposite directions, land on the same body. Kin Health raised $9 million to build an AI notetaker for patient-doctor appointments, premised on the idea that patients deserve a record of what happened in the room. The Atlantic's piece on AI wearable surveillance asks the same question from the other direction: what do you do when everyone else has a record of you too, and you didn't consent to any of it?

Theory of Mind, AI Agents, and the Healthcare Gap

The medical notetaker framing is benevolent. Patients forget. Doctors are rushed. An AI summary with next steps sounds like exactly the kind of tool that should exist. But a 2026 paper on arXiv, "Does Theory of Mind Improvement Really Benefit Human-AI Interactions?" by Gong et al., found that LLMs with improved social reasoning don't automatically produce better outcomes in high-stakes human interactions. The gap between technical capability and actual care is real, and in a medical context it's not abstract. The notetaker might catch the prescription dosage. It won't catch the doctor's tone. Meanwhile, The Atlantic argues AI has broken containment, once-speculative concerns now becoming operational realities at speed.

The Economics of Patient Data and Health AI Investment

The $9M raise for Kin Health is modest by 2026 standards, but it signals something. , especially those that sit at the compliance-friendly edge of the care continuum. The notetaker doesn't diagnose. It doesn't prescribe. It just listens. That's a very careful product decision, and a very deliberate regulatory one. Whether it's empowerment or extraction depends entirely on who owns the transcript and what happens to it next.