The Palmdoc Chronicles

Like many clinicians, my mornings are compressed. Ward rounds, clinics, calls, meetings. The day starts fast, and the last thing I want is to manually scan my calendar while rushing out. I wanted my phone to simply tell me my schedule, clearly and automatically. Not just the event titles, but the start times as well. So I decided to build a Shortcut on my iPhone that would read out my appointments each morning. The problem? Apple Shortcuts is powerful, but not always intuitive. That is where ChatGPT helped. The Objective Create a Shortcut that would: Pull all calendar events for the day Exclude irrelevant entries Speak each appointment Include the start time Sound natural when read aloud In short, I wanted something that behaved like a quiet personal assistant. Where I Got Stuck Initially, I made a common mistake. I used “Get Text from Calendar Events.” It looked logical. But it flattened the data into messy text. Once that happens, you lose access to structured fields like: Event title Start time Location ChatGPT immediately pointed out the flaw. Do not convert structured data into text too early. Instead, iterate through the events properly. This single correction saved me a lot…

Patients are saying something the medical system should not ignore. Some credit AI tools like ChatGPT for flagging danger when humans did not. Not because AI made a diagnosis, but because it pushed them to act when the system stalled. In these stories, people pasted symptoms, lab results, or scan reports into a chatbot. The AI highlighted red flags and urged urgent care. Clinicians later confirmed serious disease. The value was speed, pattern recognition, and persistence. Not certainty. This does not make AI a doctor. It makes it a gap-filler. Short consults, fragmented records, delayed follow-up. These are real failures. A language model that reads everything and responds instantly can change the timeline. In emergencies, that matters. Doctors should be skeptical but not dismissive. Yes, AI can hallucinate. Yes, it lacks examination and accountability. But it also listens without interruption and never gets tired. Ignoring that advantage is risky. The bigger risk is not patients using AI. The bigger risk is pretending they are not. Patients will use these tools anyway. The smart move is to guide them. Position AI as triage and augmentation, not diagnosis. Build pathways for rapid human review when AI raises alarms. When a patient says…

This was an experiment in vibecoding, taking medicine.com.my and rethinking how it should work for doctors. The old site was entirely WordPress powered. It functioned, but it was heavy, slow, and increasingly hard to shape without fighting plugins and themes. What surprised me was how easy the rebuild was. With a sleek vibecoded approach, the site immediately felt faster and calmer. Pages loaded quicker. The layout was cleaner. The content stood out instead of being buried under design noise. Customization turned out to be far simpler than expected. Adding a sidebar, linking social media, inserting buttons, and adjusting layout took minutes, not hours. No dependency juggling. No plugin conflicts. Just direct control over what appears on the page. This matters because doctors are busy. We assume building or modifying a website is technical and time-consuming. Vibecoding breaks that assumption. If a doctor can reason clinically, they can reason through simple code with modern tools and get usable results fast. All doctors should learn basic vibecoding. Not to replace developers, but to regain control over their own digital tools. As a side note, I’m starting to collect my vibecoded projects here: https://palmdoc.net/vibecode