1
2 Comments

Turns out, remembering your patients actually matters.

I built an AI tool for healthcare clinicians. What users actually value surprised me.
I'm a physiotherapist. I've spent 18 years treating patients, and the last two building AiDux — a clinical documentation tool for allied health professionals.
I started with a simple assumption: clinicians are drowning in notes, so the most valuable thing I could build was a better, faster way to write them.
I was wrong. Or at least, incomplete.

The problem I thought I was solving
Every allied health professional knows this moment: end of a long day, 5–10 clinical notes still pending, a head full of cases, and zero energy left.
AI scribes have appeared to help with this. They listen to a session and generate a note. Useful. But in practice, a patient is not a single note — they're a story across weeks or months. And when you need to write a referral letter, update a care plan, or just remember what actually changed between visit 2 and visit 7, most tools leave you starting from scratch.
So I built AiDux to do something more: not just document today's session, but build a longitudinal picture of the patient across visits — pain over time, function over time, key clinical events — always as a draft for the clinician to review, never as a decision.

What users actually told me
I'm in early pilot with a small group of clinicians. Completely free for now, because I'm still learning.
What I expected them to value: note quality, time saved writing.
What they actually said: "I can remember my patients better. I feel more organized. I walk into a session actually knowing where we left off."
That shift — from documentation tool to cognitive support — is something I didn't fully anticipate. And it's changing how I think about what AiDux really is.
Maybe it's less of a scribe and more of a clinical thinking assistant.

Why I'm posting here
I'm carrying this largely alone. I have clinical knowledge, enough technical ability to build and ship, and real users giving me real feedback.
What I don't have is perspective. I can't always see what I'm missing — in the product, in the positioning, in the path forward.
I'm not here to pitch. I'm here because I genuinely believe this solves something real, and I want to build it with people who can help me see it more clearly.
A few honest questions:

If you're a clinician (physio, OT, chiro, nurse, GP): does the "longitudinal memory" problem resonate? Is that what you'd actually pay for?
If you've built in regulated spaces: how did you stay useful without crossing into medical device territory too early?
If you've gone from solo founder to getting real traction: what was the thing that changed everything?

Happy to share more about the product, the stack, or the mess. Thanks for reading.

on March 13, 2026
  1. 1

    It's fascinating that the real value shifted from "faster notes" to "longitudinal memory." In healthcare, the narrative thread of a patient's recovery is so much more important than a single data point, and it sounds like AiDux is hitting a deep nerve with that "clinical thinking assistant" positioning.
    Since you've moved from a simple scribe to solving a genuine cognitive support problem for clinicians, there’s a competition where you can submit AiDux—entry is $19 and the winner gets a Tokyo trip.
    Also, the prize pool just opened at $0. Your odds are the best right now.
    How are you handling the data privacy and compliance side of things while building out that longitudinal picture—does the data stay siloed per clinician?

    1. 1

      Great question.

      We’re deliberately not designing AiDux as one big shared clinical brain where patient data gets mixed together.

      The principle is: patient context stays scoped, traceable, and governed.

      Longitudinal memory is patient-specific and clinician/clinic-scoped. We also separate documented facts, AI observations, and human clinical decisions. That distinction matters because an AI observation should not become clinical truth unless the clinician validates it.

      For future learning, I don’t think the right unit is “raw patient history.” The safer and more valuable unit is:

      signal detected → question asked → clinician response → usefulness → outcome pattern

      That can be minimized and governed much more safely.

      So yes, the data privacy/compliance side is not an afterthought. It is central to the architecture.