After two brain surgeries in 2024, a patient still faced sudden fatigue that made driving unsafe and grocery trips impossible. Diagnosed with a prolactinoma last year, the author turned to AI to map the mystery symptoms. The result: a month of consistent energy, proving that patient‑led models can match, and sometimes surpass, primary care today.

Using Claude Opus 4.8 and GPT‑5.5, the author logged symptom data, ordered blood work, and fed the results into a large‑language model. The $20 subscription to Claude Opus 4.8 proved cost‑effective. The AI surfaced nearly every hypothesis the neuroendocrinologist raised, flagging a specialized test the physician had already ordered. In twenty minutes of dialogue, the model outperformed dozens of PCP visits and reduced diagnostic uncertainty.

Eric Topol’s Deep Medicine notes that patients live amid insufficient data, time, context, and presence. The author’s workflow—track, test, analyze, experiment—provides the missing links. By keeping a longitudinal log and leveraging an always‑available model, patients gain agency they cannot get from a rushed doctor’s visit, turning vague fatigue into actionable insight for better outcomes.

Although the AI never replaced the specialist, it accelerated hypothesis generation and uncovered a test the clinician had independently ordered. The author offers a repeatable four‑step prompt and coding‑agent skill for others. Patients who face non‑debilitating symptoms now have a concrete path to use AI, avoiding costly, delayed care and improving daily life.