6 Comments
User's avatar
Michael Campbell's avatar

Fantastic post Gene. If you build the glue, it only gets stronger as the uploads are gradually replaced with low latency, validated (e.g., PIQI) data from API pipelines. Keep fighting the good fight.

manuel albarracin's avatar

However interesting the recent push that the PHR notion is receiving from the likes of OpenAI, Amazon or Anthropic, IMO it is architecturally flawed. I don’t see how the brittleness of the RLS can be solved (sounds great for financial data; health is too complex), and even if it were, the resulting information would be in the hands of wholly untrustworthy monopolists, not ours, who feed it into an also wholly unreliable LLM. I very much believe in personal agency and the PHR paradigm, but this is not the way.

Stuart Miller's avatar

Okay, I promised to return. Just published my tak,e which I know you'll like.

...so please do and comment and restack and all the other supportive things we do for each other LOL

https://haverin.substack.com/p/the-chart-is-not-the-patient?r=2vmx1l

FHIR Data guy's avatar

Incredible work!

Stuart Miller's avatar

Thanks as ever for your class leading guinea pig antics!

Im part way through a more fundamental critique of what’s missing from both the Claude and OpenAI solution approaches

And it isn’t necessarily technical more a fundamental gap.

Just like James’s Bond I will return and let you know when I hit the publish button :-)

Val Pariso's avatar

RLS is joke thanks for excellent insight. This can be solved