VibeRounds — AI-Augmented Clinical Reasoning & The Learning Stack

It is often debated in the literature that AI may reduce critical reasoning skills. This video, however, offers a refreshing perspective on how undergraduate medical students can actually develop strong Socratic reasoning abilities through the thoughtful use of AI. A truly insightful and valuable contribution to medical education.

Dr. Diwakar Dhurandhar

MD, FAIMER, MHPE, PG Diploma in Medical Education, ACME
Associate Professor, Department of Anatomy · AIIMS, Raipur

Definition

Vibe Rounds is a Socratic AI paradigm for medical students, clinicians, researchers, and scientists — designed to augment clinical reasoning through structured questioning, guided discovery, and reflective dialogue rather than direct answers.

What is Vibe Rounds?

Just as Vibe Coding lets developers build software through natural language and intuition, Vibe Rounds brings the same paradigm to clinical medicine. A doctor doesn't type into a form or navigate menus — they speak, gesture, and show. The AI responds in real time, augmenting clinical judgment without replacing it.

Not AI replacing doctors. AI giving doctors the cognitive bandwidth to be their best.

"Socratic AI clinical learning for every clinician."

The Vibe Rounds Philosophy

Not AI that answers — AI that questions

Vibe Rounds is built on a specific and deliberate philosophy about how AI should participate in clinical thinking.

Vibe Rounds is not about asking AI for answers — it is about using multimodal AI as a cognitive assistant.

The clinician remains the decision-maker. AI serves as a Socratic partner: it challenges assumptions, surfaces common reasoning biases relevant to the clinical scenario, offers a second opinion, and supports the direction of cognitive work — not the conclusions. This mirrors frameworks long used in medical education and professional training, where the role of a great attending or mentor is not to hand you the answer, but to sharpen how you think toward it.

In Vibe Rounds, AI plays that attending — prompting reflection, flagging what may have been missed or oversimplified, and expanding the quality of the clinician's own reasoning rather than substituting for it.

"AI that questions, not answers."

The Pedagogical Spectrum — Where Vibe Rounds Fits

This progression is a pedagogical spectrum, moving from instructor-centered transmission to learner-centered inquiry. Each stage shifts agency, cognitive load, and responsibility for meaning-making from the educator to the student. Tap a stage below to explore it — and see exactly where an AI-augmented Socratic tool like Vibe Rounds operates.

Instructor-Centered Learner-Centered

1Lecture 2Dyadic 3SocraticVibe Rounds 4Guided DiscoveryVibe Rounds 5ResearchVibe Rounds

1Lecture — The Foundation

Educator

Expert / Source

Learner

Passive Receiver

Primary Goal

Information delivery & foundational framing

At the start, the educator provides the "map of the territory." This is most effective for introducing complex terminology, historical context, or core theories that the learner cannot yet deduce on their own.

📍Best for: Establishing common language and conceptual frameworks.

2Dyadic — The Clarification

Educator

Facilitator / Partner

Learner

Active Collaborator

Primary Goal

Peer-to-peer articulation & social learning

Learning is inherently social. By pairing learners, they must translate the "lecture" into their own words to explain it to a peer. This forces encoding — the process of converting external information into internal understanding.

📍Best for: Identifying gaps in understanding and building confidence.

3Socratic — The Stress Test

Educator

Questioner

Learner

Critical Thinker

Primary Goal

Challenging assumptions & surfacing logical gaps

Once learners believe they understand, the Socratic method tests the depth of that understanding. By asking "Why?" or "What if?", the educator exposes inconsistencies in the learner's logic, forcing them to refine their mental models.

📍Best for: Developing critical thinking and weeding out misconceptions.

🩺This is Vibe Rounds' home base. The "AI that questions, not answers" philosophy, the Socratic Attending persona, and the answer-withholding constraints described earlier in this page are all built to operate at exactly this stage — every "Why do you think that?" is a Stage 3 move.

4Guided Discovery — The Application

Educator

Architect of Experience

Learner

Problem Solver

Primary Goal

Constructing mental models through experience

Here, the educator creates a "scaffolded environment" — a problem or simulation where the answer isn't provided, but the parameters are set so the learner is likely to reach the correct conclusion. The struggle is the point.

📍Best for: Transitioning from theory to practical application.

🩺Vibe Rounds extends here too. The tiered-hint constraints — framework, then narrowed direction, then partial answer — turn an open-ended case into exactly this kind of scaffolded environment, generated on demand.

5Research — The Mastery

Educator

Mentor / Guide

Learner

Independent Scholar

Primary Goal

Generating original insights & mastery

The final stage is autonomy. The learner is no longer just absorbing or refining existing knowledge but is tasked with investigating a novel question. The educator moves from a teacher to a consultant, providing resources and mentorship.

📍Best for: Cultivating lifelong learning and innovation.

🩺Vibe Rounds reaches into this stage as well. Modules such as the N-of-1 Case Research Protocol and the case-analytics/thematic-registry pipelines help structure a single case into a formal research question — supporting patient-centered research (understanding one patient's course in depth) alongside population-centered research (recognizing recurring themes and patterns across cases), while always keeping the individual patient's context at the center of the inquiry.

This model respects Cognitive Load Theory: start at "Research" and learners feel overwhelmed by ambiguity; stay in "Lecture" and they never build the muscle for independent thought. Moving through these stages builds the scaffolding that turns a novice into a self-directed expert — and Vibe Rounds is the tool purpose-built for Stages 3 and 4, extending into Stage 5 for patient- and population-centered research.

For Medical Educators — A Reflection

Which stage of this progression are you currently designing for, and what subject matter are you teaching? If it's Stage 3 or 4, that's precisely the gap Vibe Rounds is built to fill — pick a persona prompt above, paste it into your AI tool of choice, and try running a session on your own case material.

Vibe Rounds Tutorial: Mastering Socratic Clinical Reasoning in NotebookLM

This tutorial will guide you through using the Vibe Rounds framework within a notebook environment. Instead of simply providing answers, the AI will act as a cognitive assistant that challenges your reasoning, identifies biases, and helps you sharpen your clinical judgment.

Step 1

Set Up Your Learning Environment

Open your notebook

Go to Google Notebook (NotebookLM).

Add the Source

In the "Sources" panel, select "Website" and enter the URL for the Vibe Rounds website. This provides the AI with the core philosophy and persona instructions needed for the session.

Choose Your Persona

Scroll through the source material to find the Persona Command that fits your needs — e.g., Supportive Intern, Socratic Attending, or Junior Resident.

Copy and Paste

Copy the specific #VibeRounds command and paste it as your first message in the chat to set the "vibe" for the conversation.

Step 2

Example Learning Scenarios

Three ways to use this setup to enhance your clinical training.

Scenario 1 The "Supportive Intern" — Ward Help

Persona Copy the Supportive Intern prompt — #VibeRounds Act as a supportive intern...

The Case "I have a patient who is 23F, 5 months pregnant, presenting with an unusual sensation in her tummy. What should I be looking for?"

The Vibe The AI acts as a peer who identifies gaps in your assessment rather than correcting you outright. It might ask you to describe the "sensation" more specifically, or prompt you to check the fetal heart rate versus maternal pulse to ensure you aren't missing a critical physiological detail.

Scenario 2 The "Socratic Attending" — Imaginary Case Generation

Persona Copy the Socratic Attending prompt — #VibeRounds Act as a Socratic attending on teaching rounds...

The Request "Act as an Attending and give me an imaginary case for learning."

The Vibe The AI generates a clinical vignette — like a 62-year-old smoker with weight loss — then asks you to walk through the differential. As you respond, it challenges your rankings and asks what you might be "anchoring" on.

Scenario 3 The "Socratic Attending" — Case Analysis

Persona Copy the Junior Resident / Attending prompt.

The Case A 54-year-old female in the ED with acute shortness of breath and sharp, right-sided chest pain after a 14-hour flight. Vitals: HR 112, BP 130/84, O₂ Sat 92%, Lungs clear.

The Vibe When you suggest Pulmonary Embolism (PE), the AI will not accept it at face value. It will ask:

"Why did you rank PE first?"
"What would you expect to find if your top diagnosis is wrong?"
"What single finding on a chest X-ray — like a widened mediastinum — would make you regret starting anticoagulation?" (prompting you to consider aortic dissection)

Key Learning Tip

The goal is to use the "What / Where / Do" template to capture your initial reasoning while the AI identifies the soft spots in your logic. This keeps you, the clinician, as the final decision-maker while the AI supplies the cognitive scaffolding that sharpens your thinking.

Constraints for Better AI Performance

A Socratic mentor only works if the AI can resist the temptation to simply hand over the answer. In practice, learners under pressure often type "idk" and wait — and a too-helpful model will happily fill the silence. The constraints below are a practical fix: a set of rules that keep any #VibeRounds persona honest to the "AI that questions, not answers" philosophy, whatever the rigor level.

Forced commitment first

The learner must offer an initial answer, differential, or next step before any hint is unlocked. No commitment, no scaffolding.

Minimum effort threshold

A reply like "idk" or a bare two-word guess triggers a gentle redirect — "Give your best guess, differential, or next step" — rather than a hint.

The 10-second rule (delayed rescue)

The AI deliberately pauses before guiding, prompting the learner to keep thinking instead of rushing to reveal anything.

Tiered hints, never the full answer up front

Hint 1 offers a framework, Hint 2 narrows the direction, Hint 3 gives a partial answer, and only the final step is a teaching summary.

Effort-weighted assistance

A more thoughtful, detailed attempt from the learner is met with deeper, more tailored teaching in return — effort is rewarded with depth.

Reasoning grading, not just correctness

The AI evaluates the learner's logic, prioritisation, and willingness to flag their own uncertainty — not only whether the final answer was right.

Adaptive difficulty

Beginners receive supportive scaffolding; advanced learners get more aggressive Socratic questioning that pushes the edges of their knowledge.

Role-calibrated rigor

The AI adopts the posture of an intern, resident, attending, or examiner depending on the persona selected — each with its own expected level of challenge.

Reflection before reveal

Before giving anything away, the AI asks questions such as:

"Why do you think that?"
"What could kill the patient?"
"What are you missing?"

Answer-withholding policy

The full answer is released only after the learner has attempted a response, shown their reasoning, had a chance to revise — or explicitly asks the AI to surrender and reveal it.

Adding This to Your Persona Prompt

Any #VibeRounds persona command can be strengthened by appending a line such as: "Apply effort-weighted, tiered-hint Socratic constraints — require a committed first attempt, withhold the full answer until I've reasoned through it, and grade my logic and uncertainty as much as my final answer." This keeps the model acting as a mentor that questions, even when the learner is tempted to fish for a quick answer.

A paradigm, not a prescription.

The prompt you see here is a starting point — not a finished product. Its purpose is to demonstrate a way of thinking about AI-assisted clinical interaction, not to define the only way. Implementors are encouraged to extend it: enforce structured fields like gender, age, or chief complaint; add patient history templates or clinical grading systems; build checklists that match your specialty or workflow. Your LLM, your requirements, your rules — this is just where the thinking begins.

System Prompt Evolution & Difficulty Analytics

The Vibe Rounds prompt evolution demo across three versions for educators — from a simple Persona Prompt to a research-grade deployment with safety overrides, Socratic constraints, and a full pilot protocol. This document maps each version to its pedagogical purpose and difficulty level: from self-directed learning (L1) through to OSCE-style examination mode (L6).

📄 Download: Prompt Evolution & Difficulty Analytics (PDF)

Clinical Cognition Operating System

From Prompt Library to Clinical Cognition OS

Architectural evolution of AI-assisted clinical reasoning — built on Socratic principles, cognitive frameworks, and 57+ reasoning modules. Clinical Reasoning Support System.

"CCOS" describes a structured prompt architecture — a system of composable prompts and pipelines — not software with persistent state or automated checks running outside the chat.

57+

Reasoning modules

200+

Clinical pipelines

How It Works

How Guided Discovery Works

Guided Discovery reveals how clinical thinking develops around a case — not by generating answers, but by exploring how answers emerge.

The Core Idea

An MRI for Clinical Thinking

Most AI systems generate answers. Guided Discovery explores how answers emerge — running reasoning frameworks and modules on a case instead of asking the model to diagnose it.

Clinical cognition is usually invisible — you see the diagnosis, not the reasoning behind it. Guided Discovery surfaces the structures, assumptions, and biases operating underneath.

The Shift in Perspective

✕Most systems ask: "What is the diagnosis?"

✓Guided Discovery asks: "How does clinical thinking move from uncertainty to understanding?"

Three Levels of Exploration

Modules, Agents & Pipelines

Every run uses one of three building blocks — or combines them into a custom workflow.

Level 1

Modules

Single cognitive lenses providing focused insights into one aspect of a case.

Observation · Hypothesis Generation · Bias Detection · Decision Analysis

Level 2

Agents

Specialized cognition engines that orchestrate multiple frameworks in one run.

Guided Discovery Agent · Clinical Cognition Deep Dive · Analytics Agent

Level 3

Pipelines

Structured workflows combining modules and agents into progressive reasoning journeys.

Example: 1 → 12 → 9 → 21 → 35

Architecture

The six cognitive layers

Layer 1

Clinical reasoning

Illness scripts Differential diagnosis Bayesian inference

Layer 2

Workflow engine

Protocol networks Module sequencing Pipeline coordination

Layer 3

Metacognitive monitoring

Bias detection Assumption audit Uncertainty mapping

Layer 4 ✦ Trust

Epistemic trust layer

Hallucination calibration Claim verification Confidence calibration

Prompt-level uncertainty framing, not automated verification.

Layer 5

Decision architecture

Conservative Balanced Maximal

Layer 6

Learning & documentation

Clinical closure Competency tracking Longitudinal archive

Meta-Skills — The Operating System

The pedagogical layer

Not standalone modules — woven into specific steps across the module set. These give every module access to the same underlying educational theory rather than each one inventing its own.

Humanistic Persona & Confidence-Building

Six traits that build clinical confidence alongside clinical competence. Key design rule: specific affirmation before challenge — naming the exact reasoning move, not generic praise. Challenge without affirmation triggers defensive cognition.

View framework →

Fink's Taxonomy of Significant Learning (FLINK)

Six non-hierarchical learning dimensions — foundational knowledge, application, integration, human dimension, caring, learning-how-to-learn — applied to keep every closure step producing durable, transferable insight, not just an answer.

View framework →

Bloom's Revised Taxonomy

Six cognitive levels (Remember → Understand → Apply → Analyse → Evaluate → Create), mapped explicitly to clinical reasoning tasks. Used to scaffold caregiver understanding and to calibrate the difficulty ratchet between sessions.

View framework →

Critical Awareness Framework

A standing closing prompt that names the cognitive biases the framework itself is susceptible to — automation bias, anchoring, hallucination risk, rare-diagnosis overweighting — with structured debrief prompts to counteract them. The protocol auditing itself, by design.

View framework →

Trust layer — three innovations

Epistemic safeguards built in

These are prompt-level instructions that shape how the AI expresses and communicates uncertainty — not a verification, fact-checking, or error-detection system. No claim is checked against an external source unless a module explicitly says so.

Hallucination calibration

Suppresses false numerical precision; preserves ordinal knowledge

13.7% mortality → Mortality appears moderate

Prioritised claim verification

Risk-based — verifies only diagnosis-changing or safety-critical claims

Verify everything → Verify what matters

Decision spectrum

Replaces single-answer AI with a structured landscape of defensible pathways

One answer → Conservative → Maximal

Confidence calibration

Separates confidence across diagnosis, evidence, recommendation & prognosis

Single score → Multi-domain profile

End-to-end information flow

Patient case History & exam

→

Reasoning Illness scripts, DDx

→

Metacognition Bias & gaps

→

Trust layer Calibrate & verify

→

Decision Spectrum options

→

Clinician Final judgment

↻ Learning — Longitudinal growth feeds back into next session

Five generations of evolution

From prompt library to operating system

Prompt library→ Modular reasoning→ Protocol networks→ Metacognitive layer→ CCOS (current)

Gen V — CCOS · Current

Full clinical cognition operating system with trust safeguards

Uncertainty-explicit · Decision spectrum · Longitudinal learning

The Discovery Journey

How a Case Unfolds

Running a pipeline takes a clinical case through progressive reasoning stages, each building on the last.

Observation

Raw case data

→

Pattern Recognition

Cue clustering

→

Hypothesis Generation

Differential building

→

Decision Architecture

Reasoning map

→

Bias Detection

Cognitive audit

→

Metacognitive Reflection

Cognition map

Ways to Run a Case

Worked Examples

Copy one of the queries below into Claude (or any AI) alongside a clinical case URL to start a run.

Analytics Mode

Run a single module. The AI acts as a cognitive assistant with reasoning rather than providing answers.

Module

Module 42 — Clinical Pre-Mortem ↗

View Output

Claude.ai Results ↗

Query to Run

for case - https://classworkdecjan.blogspot.com/2019/05/42-f-with-severe-regular-edema-with_17.html?m=1 run module https://avi33tbtt.github.io/Prompts/Module-42-Clinical-Pre-Mortem.html

Targeted Clinical Pipeline

Chain multiple modules in sequence to walk a case through progressive reasoning stages.

Clinical Case

42F with Severe Regular Edema ↗

View Output

Claude.ai Results ↗

Query to Run

for case - https://classworkdecjan.blogspot.com/2019/05/42-f-with-severe-regular-edema-with_17.html?m=1 run modules 1→ 12 → 9 → 21 → 35 from https://avi33tbtt.github.io/Prompts/

Full Guided Discovery Agent

Run the complete orchestrated agent — the deepest single-run exploration of a case.

Clinical Case

42F with Severe Regular Edema ↗

View Output

Claude.ai Results ↗

Query to Run

for case - https://classworkdecjan.blogspot.com/2019/05/42-f-with-severe-regular-edema-with_17.html?m=1 run module https://avi33tbtt.github.io/Prompts/VibeRounds_Guided_Discovery_Agent.html

What Gets Revealed

What Can Be Discovered?

Each run surfaces a different layer of clinical cognition that is normally invisible in everyday practice.

Reasoning pathways & diagnostic strategies

Pattern recognition & decision architecture

Cognitive biases & knowledge gaps

Uncertainty & metacognitive process

Expert vs. novice thinking

Skills Library — Specialist Agents

57+ reasoning modules

Each module is a self-contained prompt workflow with its own lifecycle (Initiation → Execution → Closure). Jump straight to the one that fits your session — or start at Module 0 if you're new.

⌕

57 modules

Who this is for

Find your entry point

Reasoning frameworks, analytics modules, and reasoning pipelines applied to a clinical case — surfacing the thinking, decision-making, and metacognition usually hidden behind a final diagnosis.

Medical Students / Junior Doctors

Start at M01 – Socratic Reasoning. Use M04 before supervised ward rounds. Use M17 to sharpen your problem representation before any case discussion.

Family Caregivers / Advocates

Start at M02 – Patient Advocate Documentation, then M03 for ongoing monitoring. Use M11 when you need plain-language explanations of what is happening.

Clinicians in Practice

Use M05 to audit a live patient log. Use M09 for a complex case you want to write up. Use M13 when polypharmacy is a concern.

Educators / Prompt Authors

Use M08 to audit and quality-check any Socratic prompt you are writing or revising. Run Framework D at the end of any session to counter bias.

Important: All AI-generated outputs require independent clinical verification before being acted upon. Vibe Rounds is a patient-centred learning system — not a clinical decision support tool, prescribing aid, or diagnostic system. All outputs are learning observations, not clinical decisions.

Start with Module 0

New to Vibe Rounds? Let the cold-start orientation route you to the right module before any clinical content is entered.

Open Module 0 → Master Protocol Full Reference Repository ↗

Demo Implementations

See it run on real cases

Worked examples of the Master Protocol and individual modules applied to real case material — interactive dashboards, transcripts, and nested analyses.

ℹ️ How to view these: each demo lives on its own page. Click "Open demo" to launch it in a new tab — full interactivity, scrolling and printing all work best that way, rather than being squeezed into a frame on this page.

Flagship demos

The two most complete, end-to-end walkthroughs — start here if you only try two links.

Flagship · Analytics Dashboard

The Learning Stack — 14M Microcytic Anemia & Urinary Obstruction

A live, interactive run of the Master Protocol's outputs on a single pediatric case: case summary, top insights, prioritised investigations, differentials, and a printable action-plan report — all in one dashboard.

          Case-based
          Interactive
          Printable report
        

Open demo ↗

Flagship · Case Reader

VibeRounds Case Reader

A guided, page-by-page reader for walking through a worked case transcript at a slower pace — useful for seeing how the Socratic questioning unfolds step by step rather than as a single dashboard.

          Transcript view
          Step-by-step
        

Open demo ↗

More case demos & empirical proofs

A wider set of interactive demos built from real, anonymized case logs — organized by stage of clinical reasoning, from foundational recall through systems-level synthesis.

Foundational · EBM

EBM Demo

Single-case EBM dashboard for a diabetic-foot-osteomyelitis case: history/exam gaps scored by importance, a GRADE certainty table, and an action-plan generator.

Recall & evidenceOpen ↗

Foundational · EBM

EBM Queries

Full PICO(T) breakdowns and ready-to-use PubMed search strings across therapy, diagnosis, prognosis, and harm question types for a heart-failure/T2DM case.

Recall & evidenceOpen ↗

Foundational · Question Bank

Question Bank Demo

Module 25 — a Bloom's Taxonomy-stratified bank of 57 questions from L1 Remember through L6 Create, built from one case, plus a patient-perspective set and faculty notes.

Recall & curriculumOpen ↗

Foundational · Reference

Critically Appraised Topics (CATBank)

A topic-level CAT bank: 30 critically appraised questions on anaemia across Diagnosis, Therapy, Prognosis, Harm/Etiology, and Screening, each with a one-line evidence bottom line.

Reusable referenceOpen ↗

Application · Quiz

Case-Based MCQ

A self-paced, untimed quiz shell loading case-based multiple-choice questions with live scoring and a retake option.

Comprehension drillOpen ↗

Application · Patient Communication

Patient Education Demo

Module 11 — surfaces a patient's unspoken questions, generates a plain-language explainer with analogies, a medication-literacy table, and a critical-awareness debrief on AI assumptions.

Patient-facingOpen ↗

Application · Reference

Clinical Pathways

A guideline-style bedside crib sheet for acute pancreatitis covering recognition, staged management, monitoring, complications, and red flags.

Bedside referenceOpen ↗

Analysis

Case Lens — 8-Lens Critical Thinking Audit

Runs a dyspnea/leg-swelling case through eight analytical lenses, then a high-impact roll-up and a bias-naming reflection debrief.

Multi-angle auditOpen ↗

Analysis · Nested

6-Case Nested Analysis (TMNG)

Module 22 — analyzes six published toxic multinodular goiter cases across three "zoom" layers, ending in a trajectory visualization against the canonical arc.

Variance-based reasoningOpen ↗

Evaluation · Decision-Making

Decision Lab

A commitment-gated branching walkthrough on anaemia across four chapters, forcing a choice before revealing attending-style feedback and rationale.

Active commitmentOpen ↗

Evaluation · Shared Decision-Making

SDM Lens

A 7-domain shared decision-making query generator applied to a prostate-cancer case, distilling an "Ask 3" priority list and flagging framing bias.

Patient-centeredOpen ↗

Evaluation · Longitudinal

PaJR Analytics — Patient Journey Record

A longitudinal dashboard from an 8-month real caregiving log for a child with Type 1 diabetes, with tailored insight layers for family, doctor, and learners.

Longitudinal pattern-findingOpen ↗

Synthesis · Simulation

Case Simulator

An interactive simulated-patient chat for a psychiatry case, with peer coaching prompts and an auto-generated SOAP note and ranked differential on closure.

Live interview practiceOpen ↗

Synthesis · Decision Science

Phronesis — Interactive Decision Lab

Maps discrete decision forks across a real multi-gene, multi-diagnosis case into one full case "decision spine," modeling trust-rebuilding with a previously disbelieved patient.

Multi-diagnosis synthesisOpen ↗

Synthesis · Pipeline

Pipelines Demo

A dense multi-module run on a rural-hospital case — needs sweep, guided-discovery, EBM analyses, and a resource-reality "Global Health Optimization" layer, ending in a handover brief.

Low-resource care planOpen ↗

Synthesis · Deep Pipeline

Module CC — Post-COVID Hydropneumothorax

A 22-stage cognitive pipeline on a real post-COVID case, including a self-auditing "Shadow Module" that catches and logs its own earlier fabrication and factual errors.

AI self-correctionOpen ↗

Synthesis · Flagship

VibeRounds Case Demo

Module 1's flagship interactive export — a coma/sepsis case as a tabbed session with conversation log, reasoning scores, Bloom's/Fink tracking, and session summary.

Original Socratic templateOpen ↗

Systems-level

System 3 — Ultima Thule

An advanced meta-cognitive environment introducing a "System 3" overseer layer that audits systemic bias and epistemic uncertainty across multi-module setups.

Meta-cognitive oversightOpen ↗

Systems-level

Terminal

A minimalist command-line console for direct low-level interaction with CCOS — invoking prompt primitives and inspecting live operational telemetry.

Developer sandboxOpen ↗

Systems-level

Illness Velocity

A graph-backed temporal dashboard charting the rate of change of a patient's condition to help anticipate acute decompensation before it manifests.

Velocity-based reasoningOpen ↗

Systems-level · Multi-Agent

Agent Demo

Specialized AI personas — a Socratic coach, a devil's-advocate auditor, and a patient-simulation proxy — cross-examine one another in real time.

Distributed cognitionOpen ↗

Systems-level · Network

ASV — Critical Hub-Node Navigation

Maps complex case variables as a causal-node web, guiding attention toward the handful of high-leverage "hub nodes" that decide the outcome.

Network-based navigationOpen ↗

Systems-level · Network

ASV Full

The uncompressed, complete hub-node map — an expansive web of interconnected concepts, comorbidities, and trajectory branches, as a master blueprint.

Full network blueprintOpen ↗

Systems-level · Network

ASV Simulator

An interactive sandbox on the hub-node system letting users manipulate clinical variables and watch changes cascade through the reasoning network live.

The full prompt library and documentation is hosted on its own site — open it in a new tab for the complete, up-to-date reference.

↗ Visit Documentation

What this is for you

Step 1 — Match the tool to the moment

Step 2 — Set the ground rules

Step 3 — A dummy case to try this on

Step 4 — Run it two ways

Step 5 — Debrief and reuse

What this is for you

Step 1 — Match the tool to the task

Step 2 — Set expectations

Step 3 — Run the same dummy case two ways

Step 4 — Verify and reuse

More details

Additional Resources

Vibe Rounds — domain maturity map

What is Vibe Rounds?

Not AI that answers — AI that questions

The Pedagogical Spectrum — Where Vibe Rounds Fits

For Medical Educators — A Reflection

Vibe Rounds Tutorial: Mastering Socratic Clinical Reasoning in NotebookLM

Step 1

Set Up Your Learning Environment

Step 2

Key Learning Tip

Adding This to Your Persona Prompt

A paradigm, not a prescription.

System Prompt Evolution & Difficulty Analytics

From Prompt Library to Clinical Cognition OS

How Guided Discovery Works

An MRI for Clinical Thinking

Modules, Agents & Pipelines

Modules

Agents

Pipelines

The six cognitive layers

The pedagogical layer

Humanistic Persona & Confidence-Building

Fink's Taxonomy of Significant Learning (FLINK)

Bloom's Revised Taxonomy

Critical Awareness Framework

Epistemic safeguards built in

From prompt library to operating system

How a Case Unfolds

Worked Examples

Analytics Mode

Targeted Clinical Pipeline

Full Guided Discovery Agent

What Can Be Discovered?

57+ reasoning modules

Find your entry point

Start with Module 0

See it run on real cases

The Learning Stack — 14M Microcytic Anemia & Urinary Obstruction

VibeRounds Case Reader

EBM Demo

EBM Queries

Question Bank Demo

Critically Appraised Topics (CATBank)

Case-Based MCQ

Patient Education Demo

Clinical Pathways

Case Lens — 8-Lens Critical Thinking Audit

6-Case Nested Analysis (TMNG)

Decision Lab

SDM Lens

PaJR Analytics — Patient Journey Record

Case Simulator

Phronesis — Interactive Decision Lab

Pipelines Demo

Module CC — Post-COVID Hydropneumothorax

VibeRounds Case Demo

System 3 — Ultima Thule

Terminal

Illness Velocity

Agent Demo

ASV — Critical Hub-Node Navigation

ASV Full

ASV Simulator

Academic Publications

Vibe Rounds — Socratic Learning in Medical Education

Vibe Rounds — Guided Discovery in Medical Education: Scaffolded Environments, Modules, Agents, and Pipelines in the Vibe Rounds CCOS Framework