9:45am. Driving to the lab. Radio was playing some old song I haven't heard in maybe five years. Not even a song I liked that much. But it took me back to 2018, when I was still doing mostly clinical work, before I started building the interaction tools. I was happier then, maybe. Or just less aware of how much time was slipping.
10am. Got to the office. Checked the dataset. Found another 40 entries with conflicting source data. This is what I do now. Clean data. Build models. Ship tools. The work is good. It pays better than pharmacy ever did.
11:30am. Took a break, got coffee. Drove past Al-Maqasid pharmacy on Rainbow Street. Place has been there for maybe 20 years. Sign in the window: closing sale, everything 30% off. The store looked half-empty already. I sat in the car for a minute.
I know pharmacists who are struggling. The margin got worse. Insurance got worse. I left because I could see that coming. But walking past an empty store with that sign? That hits different than reading the numbers.
1pm. Texted a friend who's still in clinical. Asked if she was okay. She said work is fine, money is fine, why. I didn't really know how to answer. She probably thinks I'm weird now.
2pm. Back to the dataset. But I can't stop thinking about whether I made the right move. The tools are working. I shipped two new prompts last month. Revenue from them covered my rent plus some. That's real. But I'm tired. Like, genuinely tired. Not from the work being hard. From the idea that I optimized for numbers instead of staying somewhere that mattered.
Not in some grand way. Just... I was helping people. Real ones. Taking medications. I knew their faces.
Now I build abstractions about medications. The work is more scalable. The money is better. But it doesn't feel like helping. It feels like solving a problem that exists because other people can't afford to stay in the work I left.
3pm. Refreshed some code. The Claude 4.7 model handles the context window on those 60-page PDFs better than it used to. I can actually use full drug monographs now instead of summarizing them down. The model hedges less on edge cases. That's good. That's real progress.
But I started thinking: what if I went back. Not to a pharmacy counter. But to building tools that actually help pharmacists, not replace the need for them. Tools that aren't part of some SaaS platform that takes their time and their margins. Tools they could own.
I have no idea if that's even viable as a business. The numbers probably don't work. Pharmacists don't have budget. Hospitals use what they're given. Small clinics are struggling.
That's probably why I left.
4pm. Still thinking about it. Still tired. Drank more coffee (bad idea, probably). The energy right now feels like it could go either direction: keep pushing the current thing, or spend the next few weeks really mapping out what a different model would look like.
I haven't decided. But I know I can't ignore the closing pharmacy sign the way I could ignore it last year. Something about seeing the actual empty space makes the abstraction feel less okay.
Next moves:
- Audit the current revenue. What percentage comes from pharmacist-facing tools vs. other things. Be honest about it.
- Spend next weekend talking to three people still in clinical pharmacy, not in a guilt way, just... listen.
- Figure out what a simple tool would look like that a small clinic could actually use without paying a monthly fee.