I'm an M2 at UCLA finishing up my neurology clerkship. I came into med school with a CS background — 4 years in a BCI lab, 10 publications, the whole thing. So when everyone started talking about AI in medicine, I was neither cynical nor uncritical. I've been building with this stuff. I know what it can and can't do.
Here's what's actually in my stack right now, and what I think is worth your time.
What I Use Every Day
Anki + FSRS Not AI in the flashy sense, but FSRS (the new spaced repetition algorithm) is genuinely better than SM-2. If you're still on the default Anki scheduler, switch. The FSRS Helper add-on also lets you postpone cards when you're pre-call without tanking your retention stats.
YouTube Still the best. Ninja Nerd for mechanisms, Boards and Beyond for high-yield. No LLM has replaced a good 20-minute video for building a mental model.
Claude for concept clarification When I'm confused about something on rounds or a UWorld explanation isn't clicking, I'll ask Claude to explain it differently. Not as a source of truth — I always verify — but as a thinking partner. It's better at "explain this like I'm building a mental model" than any textbook.
What I'm Watching
Open Evidence Clinical Q&A grounded in medical literature with citations. Haven't used it extensively yet but it's getting traction among residents. The citation-grounding is the right approach — hallucinated drug interactions are not a joke.
LLMs for differential generation Interesting research space. The concern isn't that AI gets the differential wrong — it's that it's confidently wrong in ways that are hard to catch when you're tired at 6 AM pre-rounds. Still needs a lot of work before it's clinically useful.
What I'm Building
My UWorld tracker DB logs every question I've done, but right now it's just a spreadsheet with a web UI. What I want: automatic tagging of weak topics → surfaces the right First Aid section → shows me which Anki cards to prioritize before my next shelf. The data is there. The pipeline isn't.
That's the gap I keep seeing: the tools exist in pieces. Nobody has connected them yet in a way that fits how a third-year actually studies — in 20-minute windows, between prerounds and notes, on a phone.
The Honest Take
AI isn't going to save you on the wards right now. What it will do is compress the time between "I don't understand this" and "I have a working mental model." That's real. That's worth something when you're post-call and need to prep for an attending's pimping session in 8 hours.
The tools I trust are the ones that show their work. Everything else I treat as a starting point, not an answer.
I'm an MD candidate at UCLA DGSOM and CS grad from UC Irvine. I write about building at the intersection of medicine and technology.