Started with standard ChatGPT for clinicians asking for a differential for a GI bleed patient. Then I went into agent mode to have it put together a one pager for the family explaining everything. Of course, this is not a real patient.

This example does not need agent mode at all. And OpenEvidence already can do this really well.

Soooo should I buy MSFT stock at this recent pullback?

Is this the first time you've used an LLM?

🤯 We knew this was coming. Just wow

Chatgpt is great for referencing during step prep and board review (cross check facts!)

love that use case. The big Q is whether step exams need to be rethought overall.

The USMLE recall scandal of 2024 from Asia proves they do! I feel AI will be instrumental in that reform.

Open evidence has been producing very solid patient facing messages and handouts. Does this seem significantly higher quality?

Sponsored by openai

Why did you use agent mode for this? It can do it in standard

Oh dude that’s fucking awesome and huge I like the idea of the write up and document to the family on why doctors think this is what’s happening, very smart Unfortunately in the day and age of no one trusting doctors and most people thinking AI knows better, they’re much more likely to trust you even if you have the best bedside manor on earth

we can earn it back! That’s my life goal.

It will happen one day man 🙏🏻

eh, i don't see a huge difference between OE and oai for clinicians, yet.

@jeremyhynh you can now use agents and deep research… that alone is wild

You don't need agent mode to put together a one pager, like not at all needed to do that.

Did this myself it’s actually pretty incredible! Gives great understandable information for patients and honestly helps me learn.

What about getting context on your patients data. We have flirted with the idea of rolling out Solo for clinicians. If you had context on your patients, sleep, diet, exercise, supplements and daily wellbeing would that be helpful when making a diagnosis?

Agent mode? Worked for 3 minutes? My man, @openevidence can do this for you in 3 seconds.

curious what impressed you? the answer quality or the flow?