How GPT-5 helped immunologist Derya Unutmaz solve a 3-year-old mystery

Source: OpenAI Date: 2026-06-23 URL: https://openai.com/index/gpt-5-immunology-mystery

Summary

OpenAI case study profiling immunologist Derya Unutmaz using GPT-5 to work through a biological puzzle that had resisted resolution for three years. Unutmaz is a researcher known for high-dimensional immune system work; the framing is that GPT-5’s ability to synthesize across a broad literature and iterate rapidly on hypotheses compressed the timeline on a specific mechanistic question in immunology. (Source page returned 403 to direct fetch; this is a title-grounded summary from the announcement metadata.)

Implications

• Model capability: Individual researcher stories are OpenAI’s preferred format for illustrating frontier-model capability in science — they sidestep the benchmark critique by anchoring to a named expert and a concrete result. The 3-year framing is the key signal: if accurate, it means GPT-5 is compressing experimental iteration cycles in domains where the bottleneck is synthesis and hypothesis generation, not wet-lab throughput.
• Voices/power dynamics: Case studies featuring named scientists function as social proof for academic adoption. Watch whether this spawns a pattern of labs publishing “GPT-5 solved our X” stories — that would signal organized adoption strategy, not just opportunistic use.
• Agent landscape: Expert-in-the-loop science workflows — researcher poses question, model synthesizes literature, researcher evaluates — are structurally similar to the agent patterns being built in software development. The immunology context is a useful stress test because the domain is high-stakes and the evaluation criteria are rigorous.