Thin file*
chartmypath.io/AICookedMyDegree
MBBS
verdict slip · printed 1 Aug 2026 · sources on every line
HONEST VERDICT: the evidence file on this one is thinner than we'd like. 6 rows held up to sourcing: we print those and nothing else.
how to read this slip
stays humanthe evidence says this stays human led
with AIdone faster with AI in the loop: still needs you
shiftingthe evidence says AI already reaches this one
no data yetno source covers it yet, so we don't guess
STAYS HUMAN2
with AI2
shifting2
tasks charted · feeds 1 jobs6
of 6 charted tasks: 2 stay human · 2 run with AI in the loop · 2 are shifting. counts only: we don't do scores.
General Internal Medicine Physicians[feeds: source]
Analyze records, reports, test results, or examination information to diagnose medical condition of patient.with AI
Eloundou et al., "GPTs are GPTs" (2023-03, arXiv:2303.10130) · dated 2023-03.
open source ↗Treat internal disorders, such as hypertension, heart disease, diabetes, or problems of the lung, brain, kidney, or gastrointestinal tract.stays human
Eloundou et al., "GPTs are GPTs" (2023-03, arXiv:2303.10130) · dated 2023-03.
open source ↗Prescribe or administer medication, therapy, and other specialized medical care to treat or prevent illness, disease, or injury.stays human
Eloundou et al., "GPTs are GPTs" (2023-03, arXiv:2303.10130) · dated 2023-03.
open source ↗Explain procedures and discuss test results or prescribed treatments with patients.shifting
Eloundou et al., "GPTs are GPTs" (2023-03, arXiv:2303.10130) · dated 2023-03.
open source ↗Collect, record, and maintain patient information, such as medical history, reports, or examination results.with AI
Eloundou et al., "GPTs are GPTs" (2023-03, arXiv:2303.10130) · dated 2023-03.
open source ↗Refer patient to medical specialist or other practitioner when necessary.shifting
Eloundou et al., "GPTs are GPTs" (2023-03, arXiv:2303.10130) · dated 2023-03.
open source ↗the delta: what moves your picture
#1 Directing AI to draft a first-pass visit note from the encounter, then checking every documented finding against your own exam and the payer's documentation requirement before it's signed
In MBBS's own task ledger, General Internal Medicine Physicians' "Collect, record, and maintain patient information, such as medical history,…" rates augmented; the one matched row here rates augmented. Checking every documented finding against your own exam is what keeps a visit note something you'd sign your name to. first step: CMS: Evaluation and Management Services Guide · free · ~25 min
#2 Using AI to draft a plain-language explanation of a procedure, condition, or medication for a patient or family, then checking it against the actual clinical facts before handing it over
Among MBBS's matched ledger rows, General Internal Medicine Physicians' "Explain procedures and discuss test results or prescribed treatments with…" rates exposed; the one matched row here rates exposed. Checking a plain-language explanation against the actual clinical facts before handing it to a patient is what keeps it trustworthy, not just readable. first step: AHRQ: Health Literacy Universal Precautions Toolkit (free, official) · free · ~20 min
#3 Using AI as a second differential-diagnosis generator against a patient's records and results, then checking each suggested possibility against your own exam findings and the actual test data before it changes the diagnosis
Looking at the roles MBBS feeds into, General Internal Medicine Physicians' "Analyze records, reports, test results, or examination information to diagnose…" rates augmented; the one matched row here rates augmented. Checking each AI-suggested possibility against your own exam findings and the actual test data is what keeps a diagnosis yours to sign, not the tool's. first step: National Academies of Sciences: Improving Diagnosis in Health Care · free · ~25 min
* THE HONESTY BOX
thin file means exactly that: task-level research on this field is patchier than for others we cover. we could have padded this slip to look complete. we didn't. every printed line is sourced; where studies disagree, we print both. nobody can promise outcomes; anyone who does is selling something.
Task-level AI evidence on this page is largely US-scoped research; local pay and market context differ.
thin file means exactly that: task-level research on this field is patchier than for others we cover. we could have padded this slip to look complete. we didn't. every printed line is sourced; where studies disagree, we print both. nobody can promise outcomes; anyone who does is selling something.
Task-level AI evidence on this page is largely US-scoped research; local pay and market context differ.
chartmypath.io/AICookedMyDegree · no account · results private by default