Speaker: Recep Can Yavaş (Bilkent University)
Title: How Many Prompts from an Idea to a Published Paper?
Date & Time: 9 October 2026, 13:30
Place: EA 302
Abstract: Imagine taking an undeveloped research idea to a published paper as its sole author, using only a mobile phone. You cannot work at a computer, use a pen or pencil, or draw. AI agents may use computers and tools, but every stage of your work must proceed through prompting. That includes developing the idea, finding the literature, and producing proofs or data analysis, followed by drafting, reviewer responses, and manuscript revision. I do not know how many prompts this would take, but I suspect fewer than you expect.
After a brief introduction to my information theory research and results, I will describe the workflow I am developing. Agents and subagents share a folder and write documents for one another, reducing the coordination I have to do myself. I give direction and make decisions when needed, still more often than I would like. Detailed protocols aim to convey my research standards, writing taste, and preferences for proofs and exposition. AI can already rewrite a paragraph from my 2021 paper without my noticing. The harder question is whether such agents can also make research decisions as well as the researcher.
Bio: Recep Can Yavas received the B.S. degree in electrical engineering from Bilkent University, Ankara, Turkey, in 2016. He received the M.S. and Ph.D. degrees in electrical engineering from California Institute of Technology (Caltech) in 2017 and 2023, respectively. He was a research fellow at CNRS@CREATE, Singapore, between 2022 and 2024. He was with the Department of Computer Science at National University of Singapore as a research fellow. Since January 2026, he has been an Assistant Professor with the Department of Electrical and Electronics Engineering at Bilkent University, Ankara, Turkey. His research in information theory has been supported by the TÜBİTAK 2232-B International Fellowship for Early Stage Researchers. His research interests include information theory, communications, and multi-armed bandits.