About the Creator
I'm David Corteville. I'm a non-invasive cardiologist, and I write software in my free time, mostly because the tools I wanted didn't exist.
What this podcast did
This is the real reason the site exists, so it's worth saying up front.
For a long time I listened to five people talk every week about exponentials, agents, and the distance between what's possible now and what everyone still assumes isn't. It was something I enjoyed on a drive. At some point it stopped being that and turned into a list of things I should probably be doing.
I bought a Mac Studio because of this podcast. I set up an OpenClaw workspace and a group of Hermes agents that go through my own notes and projects every day now. I started writing code at night and on weekends, in the gaps around clinic, and I haven't really stopped since.
None of that would have happened five years ago. I'm a doctor. If you had asked me then, I'd have told you I wasn't the sort of person who builds things. The panel changed my mind about that without ever knowing they'd done it.
So this is a thank-you, with a scoreboard attached. "They're going to say physics is cooked again, I can feel it" seemed like the kind of prediction that ought to be worth points.
It's a fan project. I'm not affiliated with the show, nobody asked me to make it, and nothing here is official.
Projects
ResectaSep
A catheter-based device concept for taking out subaortic tissue without opening the chest.
Hypertrophic cardiomyopathy and TAVR-related outflow obstruction come down to the same problem, which is that there's too much muscle in the way. Today you either do a surgical myectomy, which works but is open-heart surgery, or an alcohol ablation, which is faster but imprecise and can't be taken back. There isn't a percutaneous, image-guided tool that lets you remove a controlled amount and stop where you meant to.
That's what I've been working on. Femoral access, 3D echo and fluoroscopy to define the resection volume, a cutting head that captures the tissue instead of letting it go downstream, and gradient measurement as you go, so you can see the result before you're finished.
It's in development. Not approved, and not in patients.
AI-driven ultrasound imaging
Ultrasound is the imaging I use most, and it's also the one where the picture depends most on who happens to be holding the probe. I think that's a solvable problem now in a way it wasn't a few years ago. I have claims drafted and figures done, and I'd rather not say more than that yet.
Open the Doors
A live music booking platform. Still being built.
I'm going to be vague about this one. The mechanism is the whole product, and I'd rather not hand it over before it ships. What I will say is that it involves real money moving between strangers for a show that hasn't happened yet, which turned out to be a much harder problem than any of the medical software. That's why it's taking a while.
Deoxylocks
A physician-formulated cosmetic scalp hydrogel with 5% 2-deoxy-D-ribose. Clear, odorless, once a day.
This is the one I point to when someone asks what AI actually changed for me, because I had no business building it.
It started with a paper. Anjum and colleagues, published in Frontiers in Pharmacology in 2024, out of the University of Sheffield and COMSATS University. They showed that 2-deoxy-D-ribose, an ordinary naturally occurring sugar, stimulated hair regrowth in a mouse model of androgenic alopecia. What caught my attention was the mechanism rather than the result. 2DDR upregulates VEGF and drives angiogenesis, which means new blood vessels, and blood supply to tissue is what I think about all day at work. Here it was doing something to follicles.
A few years ago I would have found that interesting and then forgotten about it. Instead I built the product, which I say without any pretense that I knew how. I had never run a patch test protocol. I had never written a Meta ad, or set up a Shopify store, or thought about what language you're allowed to use on a cosmetic label.
AI got me through all of it. Formulation and stability. Repeat-insult patch testing, and what safety substantiation a cosmetic actually needs sitting behind it. The storefront. The ad account, and then the much less enjoyable job of reading what the ad account was telling me. Every one of those is something I would have had to hire out, or more realistically never done at all.
The part I didn't expect to learn was claims discipline. We keep a written evidence map that grades every claim as stronger, moderate or speculative, plus a list of sentences we've decided we won't write. The 2DDR research is preclinical: one animal study, one species, and no human trial for 2DDR and hair loss of any kind. Being straight about that is harder than it sounds, and it matters more than any of the code does.
Vigil Dose
A drug-interaction engine. I started it after seeing too many near-misses that a better tool would have caught.
It's in private beta and I'm in no hurry to take it out of beta. A safety product that's confidently wrong does more damage than no product at all, so I'd rather be slow about it.
Privacy
Shared links can carry a ref code. It sets a 30-day cookie so I can see which players bring
friends. Nothing else is tracked.