What a few too many figs and my recent interactions with sick care have taught me!

Recently, I had a health incident that began absurdly: after eating too many dried figs, I woke the next morning with intercostal pain in my right hemidiaphragm and referred pain in my right shoulder, suggesting a possible nerve pathway.

In the meantime, I submitted the information to my favourite AI, and it suggested the combination of my GLP-1 intervention (which slows gastric emptying) and the figs—high in fibre—had likely formed an organic bezoar (a mesh of undigested plant matter) in the gut and that the best way forward was a CT scan to get a clear idea of the issue.

This made sense: pressure from the gut could push up on the right hemidiaphragm and trigger referred pain in the shoulder.

At that stage, the pain was merely uncomfortable. Because it was the weekend, I assumed it would pass naturally and decided to wait until Monday. If things didn’t improve, I would see my GP. Sadly, two nights passed and sleep was nearly impossible—lying down intensified the pain, and I woke every 1.5 hours to reset and walk. By 3–4 a.m., I had to get up; standing and walking were far less painful.

𝗠𝗼𝗻𝗱𝗮𝘆 𝗮𝘁 𝘁𝗵𝗲 𝗚𝗣

On Monday, I sat in my GP’s office discussing the AI’s diagnosis. He immediately ordered an ultrasound of the abdomen.

By Tuesday night, the pain was joined by fever, chills, and heavy sweats. My AI suggested there might be infection or inflammation, urging checks on the gallbladder and bile ducts.

The imaging company was painfully slow; almost a week later I am still awaiting their report. Fortunately, I could access the raw images through their clunky patient portal and feed them to the AI myself. Yes, you can do this just remove the patient details on the pics.

Ultrasound ruled out gallbladder issues—everything looked clear, and a comparison with images from a year earlier even earned me praise for a rejuvenated liver that no longer appeared fatty. A full blood panel same day looked good except for a markedly elevated C-reactive protein (CRP) of 71 mg/L, confirming inflammation. After another sleepless night, the GP arranged lung and abdominal X-rays.

The lungs appeared clear, but the gut showed agglomerated matter, confirming the fig ordeal. This still didn’t explain the night-time fevers and sweats. The AI kept urging a CT scan, hinting at a hidden obstruction or other missed cause.

𝗢𝗳𝗳 𝘁𝗼 𝘁𝗵𝗲 𝗘𝗺𝗲𝗿𝗴𝗲𝗻𝗰𝘆 𝗗𝗲𝗽𝗮𝗿𝘁𝗺𝗲𝗻𝘁

The GP finally said, “Go to hospital emergency.” Five days in, I realised I should have gone sooner. It’s always my policy to not go to ED unless absolutely necessary.

North Shore Hospital’s ED resembled a beehive that had lost its queen: people were kind but the process felt busy and inefficient. I repeated my story countless times and was told, “We don’t have a central system where we can see everything about you, so we’re handicapped.” Funny how people think Centralisation is the only answer - it’s not and I know this because this is what we are working on @symbioniq; the same data can very well reside under patient control and be shared with consent, avoiding a single vulnerable repository. But Doctors aren’t IT architects, I forgive them —they just badly need the friction removed.

𝗗𝗶𝘀𝗰𝗹𝗼𝘀𝗶𝗻𝗴 𝘁𝗵𝗲 𝗟𝗼𝗻𝗴𝗲𝘃𝗶𝘁𝘆 𝗦𝘁𝗮𝗰𝗸

Next came disclosing my experimental longevity stack. I hesitated, knowing insurers might deny coverage if any serious event could be linked to unprescribed substances. Ultimately, I had to come clean—her face was priceless. She asked who prescribed them and where I obtained them. I explained my biotech links and informal advice from leading longevity doctors, simplifying the GLP-1 story to Ozempic, as that was within her frame of reference.

In the end, the doctor kindly noted, “Patient is taking additional medications but does not wish these to be disclosed,” respecting my privacy. As she said, “You do you.” A relief.

𝗪𝗵𝘆 𝗘𝗮𝗿𝗹𝘆, 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝘃𝗲 𝗜𝗺𝗮𝗴𝗶𝗻𝗴 𝗠𝗮𝘁𝘁𝗲𝗿𝘀

This experience reinforced the value of early, definitive imaging. CT and MRI scanners are pricey and scarce, yet far more powerful at detecting conditions. In a sick-care model, access is slow. After six days and multiple inconclusive tests, a single CT in ED finally revealed the issues—plural.

Yes, the AI-identified problem with figs, slowed gastric emptying, and a possible bezoar was real. But the fever and pain weren’t from that. I had a straightforward pneumonia with fluid in the right lung, triggering the fever and CRP spike. The X-ray and ultrasound hadn’t caught it; I am still waiting for the imaging company’s reports. The CT solved everything.

A slightly more invasive scan, but when warranted it saves time and resources. I plan to include CT, MRI, and angiography in routine checks every two to three years after age 50—better informed than blind.

𝗔𝗜 𝗩𝗲𝗿𝘀𝘂𝘀 𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻

I disagreed with the ED doctor who claimed ChatGPT isn’t reliable. Current pro-grade models regularly outperform doctors on diagnostic benchmarks. The gap is cultural: AI adoption in medicine is embryonic. Clinicians, immersed in risk-averse evidence-based practice, lack time to experiment. That protects patients from fads, yet it slows progress.

𝘊𝘢𝘴𝘦 𝘪𝘯 𝘱𝘰𝘪𝘯𝘵: the doctor phoned a pharmacist to discuss drug interactions and whether any could explain glucose in my urine after a 15-hour fast or interfere with CT contrast dye. I’d already checked with AI but knew she needed confirmation from a human expert. Seems too me the solution to the AI culture shock clearly lies with Doctor augmentation by AI. 6 months ago I would have agreeed AI isn't reliable enough, today its almost no longer true if you use the correct frontier models, in 6 months I am sure it will blow everything out of the water. Mark my words.

For those of us with higher risk tolerance, running edge-case preventive regimens, the system is inadequate. We need AI tools that help doctors reach conclusions faster, explore broader knowledge, and boost efficiency. This is the gap I am hoping to close by giving everyone access to the right tools and the right privacy preserving framework.

𝗖𝗹𝗼𝘀𝗶𝗻𝗴 𝗧𝗵𝗼𝘂𝗴𝗵𝘁𝘀

I’ll skip the other “beehive” inefficiencies: misplaced clipboards, being sent to CT without an IV line, and messages lost in transit. Long ago, someone told me, “In New Zealand we don’t work people, we work WITH people.” Perhaps that means supporting inefficiencies with empathy. That sentence stuck with me. With AI automating the boring tasks, clinicians will be able to focus more of their time on compassion, listening, and continuous learning—putting back humanity where it belongs “in the pole position”.

What are your views about the correct way to use AI in Healthcare? What do you think are going to be the key challenges to scaling adoption?