GPT-5 just did what three years of conventional research couldn't — and the method the immunologist used is something any knowledge worker can steal today.
The Breakthrough (and Why It's Not Just a Science Story)
Dr. Derya Unutmaz, a leading immunologist, had been stuck on a puzzle about T cell behaviour for three years. Then he sat down with GPT-5 Pro and cracked it. OpenAI published the case study this week, and the implications stretch well beyond the lab.
The finding could accelerate research into cancer immunotherapy and autoimmune diseases — two areas where the pace of discovery has life-or-death consequences. But the real story here is how it happened.
The GPT-5 Research Workflow You Can Actually Use
Unutmaz didn't just paste a question into a chat box. He used GPT-5 Pro as a reasoning partner — feeding it years of accumulated data, asking it to surface contradictions, and iterating on hypotheses in a structured dialogue. Think of it less like a search engine and more like a very well-read colleague who never gets tired.
This approach — sometimes called "hypothesis-driven prompting" — works because GPT-5 can hold enormous context, spot non-obvious patterns across disciplines, and push back on your assumptions. It's a skill, not a shortcut. If you want to build this kind of workflow, our Loop Engineering with Claude course covers the iterative prompting structures that underpin exactly this kind of deep research session.
The broader technique of using AI to synthesise complex domain knowledge also connects directly to understanding how these models process and retrieve information — something our Decoding Language Models Tokenization course unpacks in plain English.
What This Means for Learners
If a frontier immunologist can use GPT-5 to break a three-year deadlock, the question isn't "is this tool powerful enough for my work?" — it's "am I using it powerfully enough?" The gap between average and expert AI users is almost entirely about prompting strategy and workflow design, not access.
Start small: take your own longest-standing unsolved problem at work, write out everything you know about it, and use GPT-5 (or a comparable model) as a structured thinking partner. Ask it to steelman the opposite view. Ask it what data you might be ignoring. The immunology breakthrough is a proof of concept for every knowledge worker with a stubborn problem on their desk.
