Every Learner Gets Their Own Attending: Vibe Rounds for Clinical Educators

 

Every Learner Gets Their Own Attending: Vibe Rounds for Clinical Educators

Every clinical educator knows the bottleneck. You have one attending, one consultant, one senior resident — and ten, twenty, sometimes forty learners who each need to be pushed, individually, at exactly the edge of what they already know. You can't be at every bedside. You can't pimp every student on every case. You can't sit with the one who's floundering long enough to actually rebuild their reasoning, not just hand them the diagnosis and move on.

Vibe Rounds, a project coined by Dr. Avinash Kumar Gupta in June 2026, starts from that bottleneck rather than from the technology. The pitch isn't "AI that teaches medicine." It's narrower and, for educators, more useful than that: AI that behaves like the best version of you — the attending who never just gives the answer — and that can sit with every single learner, on every single case, for as long as it takes.

The Constraint That Makes It Teaching, Not Answering

Most people's first instinct with a medical AI is to ask it for the diagnosis. Vibe Rounds is built around refusing to do that.

The system won't release a hint until the learner has committed to an initial answer. A one-word guess doesn't unlock anything either — there's a minimum effort threshold, so "idk" gets redirected, not rewarded. From there, hints arrive in tiers: a framework first, then a narrower direction, then a partial answer — with the full answer only surfacing after the learner has reasoned it out, or explicitly asked to "surrender." And critically, what gets graded isn't just whether they landed on the right diagnosis — it's the reasoning itself, and whether they flagged their own uncertainty along the way.

If you've ever run a ward round the right way — question first, silence held a beat too long, the answer withheld until the student has actually reached for it — this is that instinct, encoded as a constraint the AI can't skip past no matter how the learner tries to shortcut it.

One Attending, Infinitely Available

The practical shift for educators is availability. A persona prompt built on these constraints runs in any chat window — Claude, ChatGPT, Gemini, NotebookLM — on any case a learner pastes in. It doesn't replace you; it means the twenty learners who aren't standing next to you on rounds today still get pushed the same way the one next to you does.

Underneath the single persona sits something considerably larger: CCOS, the Clinical Cognition Operating System — 57 reasoning modules, 4 pedagogical frameworks, and 200+ pipelines, organized into six layers running from raw clinical reasoning (illness scripts, differential diagnosis, Bayesian inference) up through metacognitive monitoring and an "epistemic trust layer" that governs how confidently the AI is allowed to speak. A CCOS Builder lets you paste in a case, choose which modules to run and in what order, and get a ready-to-copy teaching prompt — sequence matters, since running a bias-check module before a reasoning module produces a different teaching pass than running it after.

For an educator, the practical version of that is simpler than it sounds: you're not writing new teaching scripts for every case. You're selecting which kind of attending your learner needs today — the one who drills differentials, the one who hunts for anchoring bias, the one who makes them defend a management plan against pushback.

Where This Fits on Your Teaching Spectrum

Vibe Rounds is explicit about where it belongs. The project maps itself onto a five-stage teaching spectrum, from instructor-centered lecture through to independent research, and places itself at Stage 3 (Socratic stress-testing), reaching into Stage 4 (Guided Discovery) — the zone where a learner already has grounding and the job shifts from delivering information to challenging assumptions and scaffolding mental models through structured struggle.

That's a deliberate, narrow claim. It is not trying to replace your foundational lectures, and it is not positioning itself as an autonomous research tool. It's built for the specific, unglamorous middle stretch of training — the stretch where most educator time actually gets spent, and where most educator time is hardest to scale.

The Tools, From an Educator's Chair

Beyond the prompt library, the project ships several free, browser-based, bring-your-own-API-key tools that run client-side:

• Case Bench — a general workbench where a learner pastes in a vignette, a discharge summary, or their own clerking notes, then works it as a live Socratic dialogue (with exam-style modes for NEET PG, USMLE, MRCP, or PLAB practice), a single-pass analytics report, or a full staged pipeline.
• Case Bench Lite — the same idea, faster and lower-friction, for a quick between-patients session.
• Case Simulator — a game-like format where a learner picks a specialty, difficulty, and setting (ED, ward, outpatient, ICU) and works a simulated patient encounter through a live console, complete with a labs dashboard, SOAP-note view, and a ranked differential that builds as the case unfolds.
• Clinical Polemos — a Healthcare Debate Bench that puts a case, a policy question, or a hospital dilemma in front of stakeholder personas (physicians, nurses, administrators, patient advocates, policymakers) who each argue strictly from their own seat. Useful for the systems-thinking and communication side of training that a differential-diagnosis drill doesn't touch.

The Part Worth Trusting: The Project Says What It Isn't

The most useful thing about Vibe Rounds, from an educator's standpoint, is that it doesn't oversell itself. The project publishes its own maturity map: AI-augmented clinical education is rated high maturity, with a working Socratic feedback loop, a six-level difficulty framework, and live deployment already demonstrated. Guided discovery research is medium maturity — the workflow is defined, but validation across multiple cases is still pending. Bedside clinical decision support is rated early stage — concept and architecture only, with no EMR or FHIR integration built.

Every page carries some version of the same line: built for personal responsibility, for learning — not for clinical use. Outputs are drafts to verify, not citable findings. That candor matters more for educators than for anyone else in this project's audience, because it's educators who are ultimately responsible for what a learner walks away believing is true.

What It Actually Changes

Not the standard of what good clinical teaching looks like — Vibe Rounds doesn't redefine that. What it changes is who gets access to it, and how often. The attending who never just gives the answer, who makes a student commit before they'll offer a nudge, who grades the reasoning and not just the guess — that's not a new pedagogy. It's the pedagogy every good educator already practices, on the one learner in front of them, when they have the time.

Vibe Rounds is an attempt to make that available to all of them, at once.

---

Vibe Rounds was coined by Dr. Avinash Kumar Gupta in June 2026. The full stack — persona prompts, the 57-module CCOS library, Case Bench, Case Simulator, Clinical Polemos, and accompanying courses — is available at avi33tbtt.github.io.