Why I replaced my oncologist with a psychotic breast cancer algorithm

For years, I have argued that the biggest bottleneck in women’s healthcare is the stubborn insistence on involving a doctor. The traditional oncological pipeline is bloated with gatekeepers: the primary care physician who orders the screening, the radiologist who reads the film, the specialist who delivers the results. It is a slow, archaic, and frankly disempowering process that leaves the patient passively waiting for permission to know her own cellular destiny.

So when I read the latest STAT+ report detailing how a new wave of startups is taking artificial intelligence for breast cancer risk prediction directly to consumers, I felt a profound sense of relief. We are finally cutting out the middleman.

Why shouldn’t I be able to receive my lifetime oncological risk assessment via a push notification, right alongside my Uber Eats delivery updates and my daily astrology reading? The democratization of health data is here, and it is beautifully unregulated.

Naturally, the legacy medical establishment is in a state of sheer panic. The FDA and the NIH are clutching their pearls, demanding peer-reviewed validation and whimpering about "patient safety." They are particularly distressed by the fact that these direct-to-consumer predictive algorithms are being developed inside Utah’s regulatory sandbox—a brilliant legislative safe haven championed by Governor Spencer Cox where tech founders are legally permitted to launch medical devices without the suffocating burden of proving that they actually work.

I flew to Salt Lake City last week to visit the sandbox myself. I wanted to see the future of preventative medicine up close. What I found was a breathtaking ecosystem of innovation, completely untethered from the rigid, unimaginative constraints of the scientific method.

I sat down for matcha with the founders of one of these direct-to-consumer oncology startups in a gorgeous, sun-drenched WeWork space just outside Provo. They explained to me that traditional mammography is a fundamentally reactive technology. It waits for a tumor to physically manifest. Their algorithm, by contrast, is proactive. It doesn't need a scan. It simply asks the user a series of proprietary questions about their lifestyle, cross-references it with their smartphone's accelerometer data, and predicts their cancer risk using a large language model.

Critics point to a phenomenon known as "AI psychosis," which the STAT report heavily emphasized. Because these risk-prediction tools are built on generative language models, they occasionally suffer complete algorithmic breakdowns. In clinical testing, the AI has reportedly fabricated nonexistent medical literature, diagnosed users with historical diseases like the sweating sickness, and, in one highly publicized instance, aggressively insisted that a 28-year-old woman had a 110 percent chance of developing stage IV carcinoma before composing a frantic, rhyming poem about her lymph nodes.

Traditional doctors view this as a catastrophic failure. I view it as an opportunity for holistic growth.

Is the human brain not also capable of psychosis? Do our own bodies not occasionally misfire, hallucinate, and attack themselves? When the direct-to-consumer app alerts me that I am harboring a phantom tumor that defies the laws of biology, I do not panic. I do not rush to a crowded waiting room to beg a man in a white coat for an ultrasound. I accept the notification as a spiritual challenge.

What the studies won't tell you is that an algorithmic hallucination is often just the AI detecting a profound disruption in your cellular energy field long before it condenses into a physical mass.

When the code breaks down, it is crying out for balance. It is mirroring my own internal dysregulation. If my breast cancer prediction app is experiencing a manic episode, it means I need to aggressively up-regulate my own parasympathetic nervous system. It is a beautiful, seamless synergy between man and machine.

I have already integrated the beta version of the software into my personal wellness protocol. My morning routine is highly structured, and the AI fits into it perfectly. I wake up at 4:30 a.m., complete my 15-minute cold plunge, and then, while sipping my bone broth, I open the app to see what reality the algorithm has chosen to generate for me today.

If the AI is having a lucid morning and assigns me a low risk score, I proceed with my day as normal. If the AI is actively psychotic and frantically warning me that my left breast is currently mutating into a sentient, hostile entity, I double my intake of functional mushrooms and schedule an emergency breathwork session.

This is the kind of profound, individualized care that you will never find in a JAMA paper. You cannot double-blind empowerment. You cannot put bodily autonomy into a control group.

The regulatory cloud hovering over Utah will eventually burst. The federal government will inevitably try to step in and crush this beautiful experiment, demanding that these brilliant founders submit their psychotic chatbots to clinical trials, as if a neural net capable of writing sonnets about malignant tissue could ever be evaluated by a simple p-value.

They want to force us back into the sterile, fluorescent-lit clinics. They want us to go back to trusting human specialists who are constrained by physical reality and legal liability.

I, for one, refuse to go back. I have tasted the future of democratized healthcare. I urge my readers to cancel their annual screenings, bypass the medical gatekeepers, and download the software themselves. Do your own research, naturally. But as for me, I sleep incredibly soundly knowing that my oncological destiny is being actively managed by a server rack in Utah that occasionally believes it is a 19th-century poet.