AI-generated personal software could let people create health tools suited to their needs: a timeline of lab results, an accessible interface in a preferred language, or a way to organize information before appointments. But a custom interface is only useful if it can access records scattered across hospitals, labs, paper files, and messages. The central argument is that patients need dependable infrastructure for using their data across different applications, so they can change tools without rebuilding their medical history. A lab-results timeline, for example, should retain each result’s date, units, lab-provided reference range, and origin.
That infrastructure must preserve where information came from and distinguish clinical records from patient-entered details and AI-generated explanations. It should let users trace summaries to their underlying records and learn when results are corrected. Permissions should be limited by data type and duration, visible to patients, and revocable; systems should also record access and minimize what applications receive, while recognizing that revocation cannot erase downloaded copies. Secure connections and clear limits matter, and authentic records do not make an application’s medical interpretation reliable. Merdra’s proposed role is to support many patient-facing tools rather than require everyone to use its own. A practical early test would be whether an independent app can retrieve authorized results with their context, respond to updates, and stop accessing data when permission ends.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.