How an AI-Savvy Founder Used Chatbots, Wearables and Second Opinions to Navigate Cancer

When Conno Christou talks about the year that changed his life, he sounds less like a patient describing a medical ordeal and more like a founder dissecting a product failure. The 35-year-old entrepreneur, who had spent years optimizing sleep, supplements, biomarkers and recovery, learned that even the most disciplined health routine cannot prevent every catastrophe. What it can do, in his case, is buy time, sharpen judgment and help a patient navigate a system that often moves too slowly for rare disease.

Christou’s story began with something that seemed almost routine. After a workout, his arm swelled. A doctor later found blood clots, and surgery was scheduled. But pre-operative testing led to an unexpected detour: an image revealed a large mass behind his sternum. A biopsy soon confirmed a rare and aggressive non-Hodgkin’s lymphoma. What followed was a rapid crash course in oncology, a stack of conflicting medical opinions and a highly personal experiment in using AI as a decision-support tool during treatment.

His experience is not a case study in replacing doctors with machines. It is something more complicated, and perhaps more relevant: a look at how a well-informed patient used chatbots, wearable data and relentless self-advocacy to ask better questions, weigh risk more intelligently and avoid a potentially unnecessary treatment at the end of therapy. In a year when millions of people are starting to turn to AI for health advice, Christou’s story shows both the promise and the limits of that shift.

The shock diagnosis that came from a routine complication

Christou had spent years behaving like a person determined to remove luck from the equation. He tracked sleep with a Whoop band and an Oura ring. He had annual bloodwork drawn and re-drawn, following the habits of longevity enthusiasts and researchers. He watched biomarkers over time and tried to align his routines with what he understood about circadian rhythm, exercise, protein intake and recovery.

That level of self-monitoring had become part of his identity. By the time he was 35 and building his second company, he says he was as attentive to the latest health research as almost anyone around him. His most recent checkup, in 2025, looked reassuring. By his account, everything appeared green.

Then his body gave him a warning he could not have predicted. After a workout, his arm became swollen. He assumed it was something minor. A week passed before he saw a doctor, who identified two blood clots in his veins and scheduled surgery.

That plan changed abruptly during pre-op testing. According to Christou, a physician returned with unexpected findings: a mass measuring 11 by 11 by 8 centimeters located behind the sternum. That discovery triggered a biopsy, and the biopsy led to a diagnosis he had never considered possible.

He had an aggressive form of non-Hodgkin’s lymphoma. The disease was rare enough to make ordinary comparison difficult, and the cause, he later learned, was tied to a random genetic mutation rather than lifestyle or stress. In his case, the tumor had likely been growing for only about three months. Had the timing been different by just a few more weeks, it could have progressed to stage four.

Christou describes the outcome with the kind of irony that often appears in stories about early detection: he was fortunate precisely because he was already in the system for a different problem. The mass was found only because he went in for clot-related care.

Two doctors, two treatment plans, one major decision

Once the diagnosis was confirmed, Christou encountered a problem familiar to many patients with serious illness: the medical system offered expertise, but not always consensus.

His first oncologist, a respected specialist, advised the less intensive of two chemotherapy approaches. Christou arranged for treatment to begin within days. But the night before the first infusion, he sought a second opinion.

The second oncologist recommended the more aggressive route. In this protocol, treatment would require continuous in-hospital infusions, repeated every three weeks across six months. The recommendation was not generic; it was based on the specifics of his pathology and presentation. The less aggressive regimen, he was told, offered roughly a 60% success rate for his case. The stronger option raised the odds to about 85%.

That divergence forced Christou into an all-too-common but often invisible patient task: deciding which expert to trust when qualified clinicians disagree.

“You hear many things,” Christou said. “You don’t have to follow the first advice.”

He ultimately gathered 12 opinions in total, leaning on his professional network and contacting hematologists and oncologists in the United States and abroad. The result, he says, was overwhelming: 11 of the 12 favored the harder regimen. He chose it.

For Christou, the choice did not feel heroic. It felt procedural. He was already used to making decisions from data, and this one, in his view, demanded the same discipline. He had no appetite for wishful thinking. The stakes were too high.

Why the second opinion mattered so much

The contrast between the two recommendations underscores a broader reality in cancer care: different physicians can reach different conclusions even when they are both competent and well intentioned. Pathology, disease aggressiveness, patient age, anticipated tolerance, location of the tumor and the balance of cure versus toxicity all shape the final recommendation.

For rare cancers in particular, evidence can be thinner and clinical judgment more variable. That is where second opinions become more than a formality. They can materially change the treatment path.

Christou’s case also illustrates how patients increasingly function as their own project managers. They request copies of scans, collect records, compare protocols, and pull in outside experts. In high-stakes medicine, that extra layer of effort can matter just as much as any single test result.

A patient’s six-month experiment in data-driven chemotherapy

Christou approached chemotherapy like a long operating cycle rather than a one-time crisis. He broke the process into smaller units, treating each week as a set of observable conditions rather than an undifferentiated blur. That mindset, he says, helped him maintain control when so much else felt out of control.

He also drew on something from his past: mandatory military service in Cyprus, which he completed at age 18. He likens his mindset during treatment to that of a soldier following orders through an extended campaign. The goal was not to fantasize about the future, but to manage the current phase and get through the next one.

Throughout treatment, he kept wearing his Whoop band and found that it often tracked the periods when his immune system was expected to bottom out. In some cases, it warned him before symptoms fully arrived. That kind of correlation did not make him invincible; it made him more observant.

He also kept a detailed symptom log using voice transcription. Side effects, medication changes, timing of fatigue, reactions to supportive drugs and other small shifts were all recorded. Over time, he narrowed his attention to three main variables: sleep, nutrition and psychology. Of those three, he believes the last mattered most.

Christou said mindset was the factor that moved the needle most during treatment, and that he never spent time asking why the illness had happened to him.

That view may sound stoic, even severe, but it is consistent with how many founders approach adversity. Problems are not solved by despair; they are solved by information, process and momentum. In his telling, the treatment period became an exercise in endurance with feedback loops.

Where AI entered the picture

As treatment progressed, Christou began feeding a growing archive of personal health data into Claude, the AI assistant from Anthropic. The inputs included blood tests, scan reports, wearable data and journal entries. He did not treat the model as an oracle. He used it as an analytical layer.

That distinction matters. Consumer chatbots have become a common first stop for medical questions, but clinicians warn that the tools can be confidently wrong, incomplete or overly generic. A March public opinion poll found that about one-third of American adults now use chatbots for health-related information and advice, a sign of how quickly this behavior is spreading.

Researchers and hospital leaders have urged caution. General-purpose systems are not the same as validated medical software, and they have not been tested as rigorously as clinical decision tools. For a person with a complex or rare condition, an error can be more than a nuisance; it can change treatment.

Christou does not dispute any of that. He is explicit that the chatbot did not substitute for medical care.

He said the model did not replace his doctors, but helped him frame better questions and understand what to ask next.

That may be the most defensible use case for AI in medicine today: not diagnosis by machine, but better preparation by patient. For someone with a disease that many oncologists may encounter only once a year, the ability to query a system trained on vast amounts of medical text can be valuable, especially when paired with human expertise.

Why rare disease creates an AI advantage

Common illnesses are often well covered by standard guidelines. Rare conditions are different. The treatment literature may be fragmented. The number of specialists familiar with a particular subtype can be limited. Imaging results can be ambiguous. Follow-up decisions may depend on details that are easy to miss under time pressure.

In those situations, a language model can function like a research assistant, surfacing possibilities that might otherwise be overlooked. It does not validate those possibilities. But it can help a patient, caregiver or clinician know which questions deserve attention.

Christou says that a model trained on broad medical literature was far more useful than a conventional web search. That difference is especially important when the condition is unfamiliar and the relevant evidence is scattered across papers, guidelines and case studies rather than appearing in one neat summary page.

The moment AI helped avoid more treatment

The most consequential AI-assisted decision in Christou’s story came at the end of therapy. His final PET scan was unclear. PET imaging can be extremely useful in cancer care because it helps identify active disease, but post-treatment scans are also notorious for ambiguity. In Christou’s case, the uncertainty opened the door to a potentially alarming next step: more therapy, including possible radiotherapy near the heart and lungs.

He was not comfortable accepting that conclusion without checking it carefully. He read further and found that, for this particular lymphoma, end-of-treatment PET scans can produce false positives at a striking rate. He says the figure he found was around 60%, a number that startled him enough to dig deeper.

He then uploaded his three PET scans and an MRI into Claude. The model highlighted a known phenomenon that can confuse imaging in younger patients recovering from this type of lymphoma: thymus rebound. After chemotherapy, the thymus can reactivate in people under 40 and may appear on scans as if active disease has returned.

Based on his age and the characteristics of the scan, the model suggested that this explanation was highly plausible. Christou sought additional human opinions again, and a fourth physician confirmed it: the scan pattern was consistent with thymus rebound rather than active cancer. The proposed radiotherapy was not necessary.

That outcome mattered not just because it spared him another round of treatment, but because it showed how AI can alter the burden of proof. Instead of accepting an alarming interpretation at face value, he was able to challenge it with context, literature and follow-up review.

What the timeline looks like

Christou’s course from symptom to clearance moved fast, but each stage mattered. The sequence below shows how quickly a seemingly unrelated issue escalated into a life-changing diagnosis and then into a high-stakes treatment decision.

What his experience says about the modern patient

Christou’s experience points to a broader transformation in healthcare: patients now have access to tools that can translate, summarize and compare information at a speed that was impossible only a few years ago. That does not make the system simpler. It makes the patient’s role more demanding.

To use AI well in medicine, a person still needs judgment about what to ask, what to trust and what to verify with a clinician. In other words, the technology does not remove the need for expertise. It increases the premium on knowing how to use it.

In Christou’s case, the AI workflow seems to have offered three concrete advantages:

• It organized a large volume of personal health data into something searchable and interpretable.
• It surfaced medical concepts, such as thymus rebound, that helped explain ambiguous imaging.
• It prompted him to seek additional human opinions rather than accepting a single conclusion too quickly.

That combination of machine assistance and human verification is likely to define the near-term future of consumer medical AI. The chatbot is not the last word. It is the first pass.

Why experts remain skeptical

Medical professionals are right to be wary of general-purpose AI tools in clinical settings. These systems can miss nuance, misread context or oversimplify risk. They may also reflect the statistical patterns of their training data rather than the lived complexity of an individual patient. In healthcare, a plausible-sounding answer is not enough.

That is why Christou’s story should be read carefully. It is not evidence that AI can safely manage cancer care on its own. It is evidence that a savvy patient, working with qualified physicians, can use AI to reduce confusion and improve decision-making.

The difference is crucial. One version promises replacement. The other promises augmentation.

How illness changed the founder

Christou built Keragon, an AI-powered platform designed to automate administrative tasks for medical practices, before he became a cancer patient. The experience of going through treatment gave that work a more personal dimension. He saw the operational drag on clinicians up close: paperwork, coordination, repetitive tasks and the constant churn of administrative burden layered on top of actual care.

He also saw how side effects are often managed in layers, with one drug leading to another and another, in a cascade that can become hard for patients to track. In his view, the current model of treatment may eventually be remembered as clumsy, even wasteful.

There is a practical humility in that conclusion. Patients do not just need more medicine. They need better systems. They need clearer explanations, less fragmentation and more time for decisions that matter.

Christou says the experience also altered how he thinks about daily life. He tries to keep Sundays free. He tries to be present at lunch with friends, at home with his dog and in conversations that once might have felt secondary to work. A venture capitalist friend once told him to be happy now, advice he says became more meaningful during treatment.

The line is simple, but in the context of a life-threatening diagnosis it takes on added weight. Health optimization can become a form of deferral, an endless promise that the best life begins later. Cancer has a way of exposing the cost of that mindset.

The larger takeaway for AI and healthcare

The most striking part of Christou’s story is not that AI answered everything correctly. It didn’t. The important part is that it helped him navigate uncertainty at a moment when the stakes were enormous and the information was hard to parse.

That puts his experience in the center of a fast-growing debate. As more people use chatbots to interpret symptoms, scan results and treatment options, the question is no longer whether AI will enter the exam room in some way. It already has. The question is how responsibly it will be used, and whether patients, doctors and institutions will develop standards quickly enough to keep up.

For now, Christou’s view is pragmatic. AI did not cure him. It did not make chemotherapy easy. It did not eliminate the need for experts. But it did help him make better decisions, challenge a misleading interpretation and feel less helpless inside a system that can otherwise leave patients waiting for answers.

He is still processing what the past year means for his work and his health. Yet the conclusion he draws is unmistakable. The tools are already here, and patients are already using them. The only real question is whether the medical world will adapt to that reality fast enough.

“It’s not happening in 10 years,” Christou said. “It’s happening today.”

In an age of AI-assisted everything, that line may be the clearest diagnosis of all.

Key facts at a glance

Why the story resonates beyond one patient

Christou’s account will likely resonate for a simple reason: many people now feel that the healthcare system is too complex to navigate passively. When the diagnosis is rare, the scan is unclear, the specialist recommendation changes from one doctor to the next, and the clock is ticking, patients want more than reassurance. They want leverage.

AI, for all its flaws, can provide some of that leverage by making information more usable. But the story also reminds readers that the burden of interpretation still falls on people. The best outcome came not from blind trust in a model, but from the combination of machine assistance, repeated human review and a patient willing to push for clarity.

That may be the most important lesson of all: in medicine, as in startups, the smartest move is often not to accept the first answer. It is to keep testing until the picture becomes clear enough to act on.

superintelligencenews@gmail.com