On academic rounds in the ICU, after positing back to the new intern, in the 2nd month of their physicianhood, that their working etiologic diagnosis was flawed; while offering every OMS/MS/PGY's favored format of MCQ of probable DDx including their IMprobable dx, they again chose their initial answer. DESPITE the rounds discussion. When I corrected that it was NOT that, do you know what the response was..? "...that's not what #OpenEvidence said..." I asked them to show me the prompt they utilized to ask for OE to sift through the available literature. And it was... flawed.. to say the least.What ensued was a 5 minute papal decree of pontification. We have NOT been doing a good job as a community & profession of teaching these new and soon-to-be-clinicians to THINK. CLINICALLY_AND_CRITICALLY. If you don't have a working knowledge, a level of MASTERY, in a topic: how can you know when the LLM pulled the wrong set of studies to review? Not all literature is made equal. GENERALLY it pulls guidelines & landmark trials. But what happens when you're now visiting Esoteria..? You know, the land of esoteric questions. My mentor in my first fellowship advised me that to best prepare for fall PGY4 by reading/knowing guidelines. Not memorize them; be able to reason through them all. As an ID fellow, there were only DOZENS. But, 6 months in at our 2nd mentoring sitdown, I told him that I had gone through all the guidelines at least once, and the major ones (UTI, SSTI, bacteremia, etc) SEVERAL times as consult coverage went on. He said: "Great. Now toss that out. Guidelines are for non-specialists." So: What did he MEAN by that..? We all have to have a framework; where to start. #Guidelines allow all-comers to start the path of tx, but it's for us to be able to see where perhaps the consulting team fell off said path, or when we needed to intentionally deviate from it. The #IDSA C diff guidelines reviewed studies that mostly dealt w mod to severe CDI. They publish recs in 2018 regarding CDI removing Flagyl from the equation. The works of Zar, Nijjar-Debbiny, Stevens, and more, are largely ignored in mild disease. I mean, Zar even has a SCORING SYSTEM named for him in C diff that is largely the criteria that the CDI classification system is BUILT upon. The outcry: swift. Flagyl = viable tx for outpatient CDI. Further studies completed and more reviewed. The GI docs (ACG) publish competing guidance to the updated 2021 IDSA guidance. ACG leaves in narrow window for Flagyl in non-severe mild disease. Why the disparity? I can only speculate. ALL TO SAY: Train the AI to your patterns & do prompt design training. Because scanning title+abstract+n-value+CI+conclusion does NOT a physician make, as COVID showcased. But: unless you advise a "journal club," largely that's what AI does also. Studies ALREADY SHOW we're offloading TOO MUCH cognitive load to AI. As a Chair in #UME and faculty in #GME, it's EVIDENT. I recommend to them that they hold off on AI for A&P until they can write up 90% of their daily charting As&Ps without looking up every detail. Until then, they're just skipping the development of a #physician's true tool: clinical reasoning. That's the WHOLE POINT of the training. Give the knowledge base in UME, train them how to talk and analyze in the order we all understand, then case after case after case of patients until they know normal to abnormal and which abnormal to which abnormal, all while highlighting #empathy, #professionalism, and #physicianleadership. #meded #physicianled #AI #medtwitter #idtwitter #idxposts #FOAM #FOAMed #FOAMim #FOAMcc #FOAMid #academicmedicine #criticalcare #idcrit #AIinmedicine @DrNeilStone @DrToddLee @BradSpellberg @wwrighID @mohammad @DrDiGiorgio @djschuett @PulmCrit @pulmtoilet @IdVilchez @ABsteward @medrants @joshuadhartzell @Cortes_Penfield @IDiots_pod @JAMA_current @NEJM