Trendline: from one doctor’s problem to a live product in under two weeks

A doctor I know makes home visits. As he saw more patients, keeping track got harder: a notebook for one, a spreadsheet for another, loose lab printouts, notes in his phone. When he sat down with a patient, he wanted to see how their numbers had moved since the last visit. Instead, he was flipping between four places trying to piece it together.

He wasn’t asking for an app. He described a problem: each patient’s records are scattered, and I can’t see the trend. Working out what to build from that was my job.

Why off-the-shelf apps didn’t fit

There are plenty of health apps. They’re built for one person tracking their own steps and weight, not for someone following dozens of other people’s clinical metrics over months and explaining the trend during a visit. The mismatch wasn’t the features — it was the way the work happened. That’s usually a sign that something needs to be built rather than bought.

How it got built fast

Two things kept it to under two weeks. The first was talking to him almost every day. I’d show a rough version, he’d use it against a real patient that afternoon, and tell me what was wrong. No requirements document could have surfaced what those sessions did.

The second was using AI heavily for the parts that are just typing — schema, forms, charts, the boring plumbing. That freed the time for the part that actually mattered: getting the one screen he lived in to feel right.

The first version was ugly and did one thing: show a patient’s metric as a single line over time. That was enough for him to stop using the notebook.

Where it went

Once it worked for him, the same core turned out to fit other people. Families started using it to keep everyone’s health metrics in one place. Photographing a lab report and having the numbers read automatically removed the last bit of manual entry. None of that was planned up front — it followed from a working product in real use.

If this sounds familiar

The pattern is: you describe how the work actually happens today, and the messy parts. Whether the answer is an automation, a small tool, or a full product — and whether it’s worth building at all — is my problem to work out, not yours.