Two Courses That Teach Clinical Thinking by Refusing to Give You the Answer

Most medical training hands you conclusions: the diagnosis, the guideline, the "correct" answer at the back of the case. What it rarely shows you is the reasoning that got there — the messy, iterative process of turning a worried patient into a testable question, weighing shaky evidence, and updating your mind when the data doesn't cooperate.

Two new self-paced courses — Evidence-Based Medicine, From First Principles and Clinical Cognition, From First Principles — are built around the opposite bet: that the reasoning trace is the thing worth teaching, and that an AI tutor is more useful when it questions you than when it simply tells you what to think.

Both are part of something called VibeRounds, a broader "Clinical Cognition Operating System" built on two ideas: Socratic learning, where the AI's job is to keep asking rather than answering, and Guided Discovery, where you're pushed to attempt a real answer before anything is revealed to you. Neither course is a passive read — each lesson expects you to actually do the exercise, because the homework from one lesson becomes the raw material for the next.

Course one: learning to trust (or distrust) the evidence

Evidence-Based Medicine, From First Principles started life as a single session taught to a mixed room of medical students and people with no clinical background at all — and that mixed-audience DNA still shows. Rather than opening with statistics, it opens with a patient: taking a history, running an exam, writing a SOAP note, and noticing when a claim floating around online is actually nonsense.

From there, the nine lessons build in a deliberate arc:

• Turning a vague clinical worry into a proper, searchable PICO question
• Appraising a randomized controlled trial by hand — validity, results, applicability, plus the arithmetic behind ARR, RRR, NNT, and confidence intervals
• Reading a systematic review and meta-analysis, including forest plots, heterogeneity, and publication bias
• Making sense of diagnostic test statistics — sensitivity, specificity, likelihood ratios, and why the same test means something different in a different population
• Distinguishing prognosis and harm studies from trials, and why a cohort design is sometimes the only honest tool for the job
• Reading a clinical practice guideline the way its authors built it, GRADE and all — and knowing when to override it for a patient who doesn't fit the average
• A final statistics deep-dive aimed squarely at skeptics: p-values versus confidence intervals, p-hacking, surrogate endpoints
• One last lesson that runs an entire real, de-identified patient case through the whole pipeline, start to finish

If you're coming in with zero clinical background, there's a standalone Techie Summary that re-maps the whole course into language a technical reader already speaks — PICO as a search schema, an RCT as an A/B test, a likelihood ratio as a Bayesian update, GRADE as something like a ship/no-ship decision. There's also a companion prompt library for using an LLM as a research assistant at each stage — drafting a PICO question, building a search string, screening abstracts — governed by one non-negotiable rule: the model drafts, you verify every number and citation against the source yourself.

Course two: watching the reasoning happen, not just the diagnosis

Where the EBM course asks "can I trust this evidence?", its sibling course asks a different question: "how did I actually get from findings to a judgment in the first place?" Clinical Cognition, From First Principles is a thirteen-lesson dual-track course — built for clinicians and technical readers alike — pulled from a much larger, 57-module VibeRounds prompt library.

The first nine lessons build the core reasoning loop: making the invisible logic behind a diagnosis visible, learning the Socratic questioning method itself, building a differential from raw findings through to a probability-weighted shortlist, auditing your own reasoning for bias and failure points, translating a case for the people it gets handed off to, scaling from single-patient reasoning to population-level analytics, borrowing failure-mode analysis from engineering, and reporting confidence across several independent dimensions rather than one fuzzy score. Lesson nine runs the entire "Master Protocol" against one full case, read back as a single artifact to check whether it actually holds together.

The last four lessons take the same machinery further out: meta-cognition and self-critique, multi-agent healthcare systems and operations, precision medicine and personalization, and finally the harder-to-teach territory of clinical wisdom — telling real expertise apart from a shortcut that merely looks like it. Five shorter elective modules round things out, covering N-of-1 research, journal reading, community medicine, thematic analysis, and cross-case learning.

Why they're built as a pair

You can start with either course — they're complementary, not sequential — and several Clinical Cognition lessons link directly out to the matching EBM lesson where the two overlap. Read together, the pairing covers something medical training usually leaves implicit: not just what counts as good evidence, but how a clinician's mind is actually supposed to move once that evidence is in hand.

If either topic sounds like your kind of rabbit hole, both course indexes are open and free to work through at your own pace:

• Evidence-Based Medicine, From First Principles →
• Clinical Cognition, From First Principles →

Questions, feedback, or interested in collaborating? The courses were built by Dr. Avinash Kumar Gupta, reachable via email at avi33btt@gmail.com or on LinkedIn.