# Akash Stephen

- **Event:** [Built with Opus 4.7: a Claude Code hackathon](https://cerebralvalley.ai/e/built-with-4-7-hackathon)
- **When:** Apr 21 at 12:00 PM – Apr 27 at 2:00 AM (EDT)
- **Where:** Online
- **Team:** [Akash Stephen](https://cerebralvalley.ai/u/akashstephen)
- **GitHub:** https://github.com/akashstephen/Cortex_Demo
- **Demo video:** https://www.youtube.com/watch?v=sjnVm8lzsyk
- **Gallery:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/123

Medical students memorise twenty thousand facts. They sit through two preclinical years of flashcards, mnemonics, and question banks, and they believe — every educator believes — that this is how medicine is learned.

Then they reach the bedside.

A 67-year-old man collapses while climbing stairs. Not at rest. Only on exertion. The student knows the words aortic stenosis, syncope, exercise intolerance. They have seen the flashcard. And they freeze, because the patient does not look like the flashcard, and the brain they were trained to use has no machinery for what to do next.

The crisis in medical education is not a knowledge shortage. It is a mental-model deficiency. We have spent a century teaching students to retrieve facts. We have never seriously tried to teach them to generate understanding from mechanism. Every existing tool — Anki, UWorld, Amboss, Osmosis, Pathoma, Sketchy — is a retrieval tool. None of them teach the student to derive.

Cortex is the first tool that does.

What Cortex Is

Cortex is a mechanistic medical learning system. It teaches medicine the way physics is taught: not by memorising Newton's laws, but by throwing balls, watching them arc, and discovering that the formula is just compression of what you already feel.

The current demo is a live cardiology tutor. The student is given a parameter-bound heart — an anatomical SVG whose chambers, valves, and great vessels are wired to six physiological knobs (heart rate, contractility, preload, afterload, systemic vascular resistance, valve compliance). When a knob moves, the visual heart responds in the same frame. When the student scribbles calcium onto the aortic leaflets, the valve area shrinks live from 3.0 cm² to 0.7 cm², the leaflets stiffen, the left ventricle strains, the lungs cloud with coral haze.

Then the student picks any concept they want to learn — severe aortic stenosis, cardiogenic shock, septic shock, hypertensive emergency, the vasodilator trap, or anything else they type in — and Claude Opus 4.7 takes the wheel. It dials the sliders with smooth tweens, narrates from first principles, asks the student to predict the next move, and grades the prediction against the deterministic physics behind the visual.

The student does not watch a lecture. They turn knobs on a heart, then ask Cortex to teach them whatever they do not yet understand. 

Each interaction with the platform creates a cognitive profile of the user's understanding of the human body and the subsequent sessions are dynamically made to teach the specific errors that the user is making. 

The same architecture — a parameter-bound visual world, an agentic Opus teacher with typed tools into that world, and a clinical-grounding critic refusing claims the physics does not justify — scales to teach all of medicine on a first-principles basis, backed by the way the human brain actually forms durable knowledge: through prediction, perturbation, error, and repair, not through retrieval of rehearsed facts.

---

Markdown version of https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/123. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt