About
The full story behind Second Opinion.
The why
My fiancée was diagnosed with adenomyosis last year.
She walked into her appointment after 11 years of being told heavy painful periods were normal. She walked out with two treatment options laid in front of her: hormonal contraceptives now, hysterectomy later. There was no mention of pelvic MRI for proper staging. No mention of GnRH antagonists like relugolix or linzagolix. No mention of focused ultrasound. No reference to the April 2026 Kho Clinical Expert Series in Obstetrics and Gynecology, which lays out a full medical, interventional, and surgical pathway any OB-GYN can apply.
When I read that paper for her, I learned there were at least four evidence-based options between “try the pill” and “remove the uterus” that her specialist had not put on the table. Plus another seven quality-of-life things her doctor never mentioned: iron, magnesium, anti-inflammatory diet, TENS, pelvic floor physiotherapy, cycle-aware scheduling, and peer support. I figured all of that out by reading research and watching her, not by asking a clinician.
That gap is what this is. A tool that reads what her doctor doesn’t have time to read, and hands her the brief she should have walked into the appointment with.
The pattern
She is not the only one this is happening to.
One in three women live with adenomyosis. The diagnostic delay averages eleven years. Zero drugs have ever been developed exclusively for the condition. The whole field has left clinicians behind, and patients are paying for it.
Most patients with adenomyosis are not under-diagnosed because the disease is rare. They’re under-diagnosed because the standard initial care pathway doesn’t escalate fast enough to confirm it, and because the literature on what to do about it has moved faster than clinical practice.
How it was built
Five days, one engineer, a partner with a real diagnosis.
Built for the Cerebral Valley “Built with Opus 4.7” hackathon over five days in April 2026. The clinical knowledge corpus (52 sources, ~32,500 tokens) was assembled in parallel by three research agents reading recent literature, regulatory filings, treatment-centre directories, and ClinicalTrials.gov. Every source is real and citable.
The model is Claude Opus 4.7. The corpus stays in the prompt cache (one warm call costs about thirty-seven cents); only the patient profile is dynamic per request. The output is a strict JSON schema the frontend parses into treatment cards with hover-revealing citation pills, red-flag detection, care-centre contacts, trial matches, and a printable doctor handoff page.
Stack: Next.js 16, React 19, Tailwind v4, TypeScript, Anthropic SDK with ephemeral prompt caching, Vercel hosting. Editorial palette and typography (Fraunces serif, warm cream paper, aubergine ink) chosen against the AI-startup-dark-mode default to read as a serious women’s health publication, not a tech demo.
What’s next
Adenomyosis is the first.
The architecture (curated clinical corpus + 1M-context reasoning + structured citation-backed output + printable clinical handoff) generalizes to any chronic condition where the literature has outpaced practice. Endometriosis. PCOS. Long COVID. Fibromyalgia. EDS. MCAS. Lyme. The pattern is the same: patients dismissed by overworked specialists, evidence sitting in PubMed that nobody is synthesizing for them.
Adenomyosis is the wedge because it’s where my fiancée lives. If you’re building in this space, work in patient advocacy, or are a clinician open to feedback on the brief format, please reach out.
The builder
James Olusoga · Toronto
AI Solutions Engineer and Creative Technologist. Builds production AI systems and consumer-facing tools through TdotsSolutionsz. Open to collaborations on healthcare AI projects, particularly in under-served chronic conditions.
Reach out at dev@jamesdare.com or on LinkedIn. Source code: github.com/DareDev256/second-opinion.