It started, as most good stories in operating rooms do, with a quiet conversation between scrubs.
Dr. Vijay Muthukumaran had spent years thinking in flaps and fascicles. In Dharmapuri, where the patient who walks in at 3 AM is just as likely to be a road-traffic case as a planned reconstruction, his world was tactile — the pop of a perforator, the shimmer of an anastomosis under the loupe, the slow patience of a 0.1 mm vessel waiting for a dilator.
Then a neurosurgeon friend mentioned, almost in passing, that he’d been vibe coding something on the weekends.
The phrase stuck. Vibe coding. It sounded like the opposite of surgery — sloppy, intuitive, forgiving. A surgeon’s hands are trained to never trust the first attempt; coders, apparently, were out there shipping things at midnight on a hunch.
Vijay opened his laptop.
The first project was small and personal: a follow-up app for his post-discharge patients, because WhatsApp had become a graveyard of unanswered photos and worried mothers. He called it vjm-followup. React on the front, Firestore on the back, asia-south1 because of course. He didn’t know what half those words meant when he started. He learned them the way he’d learned microsurgery — by doing the same thing badly, then less badly, then well.
Claude Code became a strange new resident on his team. Tireless, occasionally wrong, always willing to explain why. He’d describe a problem the way he’d describe a flap to a junior — tissue here, pedicle there, donor site, constraints — and a working scaffold would appear. He started catching himself thinking in commits between cases.
The neurosurgeon turned out to be the perfect partner-in-crime. Two surgeons who already knew how to debug under pressure, now arguing about state management instead of intracranial pressure. They traded snippets the way they used to trade journal articles.
The projects multiplied. A flap library PWA, because every plastic surgeon in the country wanted one and nobody had built one that wasn’t a PowerPoint in disguise. An on-device Gemma model running through MediaPipe, because patient data had no business living on someone else’s server. A RAG pipeline over his own clinical notes. An iOS experiment with the Anthropic API, key tucked safely in an .xcconfig file like a suture buried under skin.
What surprised him most wasn’t the code. It was the realization that surgery and software shared a spine: a problem stated cleanly, a plan drawn before the first cut, a willingness to undo and redo, and the humility to know the second version is always better than the first.
The flap. The function. The follow-up.Same craft. Different scalpel.