Can AI do Theory? Workshop Cover

Can AI do Theory?

Workshop at STOC TheoryFest 2026

June 27, 2026 | Salt Lake City, Utah, USA

This workshop will explore the intersection of artificial intelligence and theoretical computer science.
Through invited talks, a panel discussion and a poster session, we hope to foster a community-wide dialogue on whether and how AI can augment our current methodologies or redefine how we approach research in theoretical computer science.


Invited Speakers

In-Person Speakers

Scott Aaronson
Scott
Aaronson

UT Austin

Mark Selke
Mark
Sellke

Harvard, Open AI

David Woodruff
David
Woodruff

CMU, Google


Virtual Speakers


Panel Discussion

The panel will feature a number of established researchers who will share their perspectives on how they see AI influencing the future of theoretical computer science.

Confirmed Panelists:

Mohammed Abouzaid
Mohammed
Abouzaid

Stanford

Yael Kalai
Yael
Kalai

MIT

Jon Kleinberg
Jon
Kleinberg

Cornell

Raghu Meka
Raghu
Meka

UCLA


Organizers

Accepted Posters


Call for Posters

We seek poster submissions that explore the intersection of artificial intelligence and theoretical computer science. Topics of interest include, but are not limited to:

These could be related to any domain of interest to theoretical computer science or mathematics.

We welcome contributions from researchers and practitioners across all backgrounds, including academia and industry. If you are uncertain whether your research fits the exact scope, we strongly encourage you to submit regardless!

Key dates:

The poster submission is closed. Accepted submissions are listed above.

Program

Time Session
08:00 - 09:00Coffee and Breakfast
09:00 - 09:15Opening Remarks
09:15 - 10:00Scott Aaronson
Are We Cooked?
Abstract: I will describe some recent case studies of using GPT 5.5 Pro to help solve open problems in quantum complexity theory, and reflect on what theoretical computer science might look like going forward.
10:00 - 10:30Prabhakar Raghavan (virtual)
Some early successes and learnings in LLM-assisted Math and Theory
10:30 - 11:00Sébastien Bubeck (virtual)
Probability, combinatorics & optimization from GPT-5 to GPT-5.5
Abstract: I will describe five moments that changed how I think about AI and mathematics, with the first one being when GPT-5 was released and the most recent one with GPT-5.5.
11:00 - 11:45David Woodruff
The AI Review Process and Beyond
Abstract: I will discuss the results of the STOC 2026 experiment, where we used inference time scaling on top of Gemini-based models (specifically advanced versions of Gemini Deep Think) to give authors pre-submission feedback, which has now been expanded and deployed at ICML and NeurIPS (see blog).

I will then discuss how this system was also used as a key verification tool used by researchers to solve open problems, refute conjectures, and generate new proofs across diverse areas in theoretical computer science, as well as other areas such as economics, optimization, and physics, as described here.
11:45 - 13:15Lunch (provided)
13:15 - 14:00Mark Sellke
Disproof of the Unit Distance Conjecture
Abstract: I will discuss the recent disproof of the unit distance conjecture due to an internal model at OpenAI.
14:00 - 14:30Carina Hong (virtual)
Frontiers of AI for Math and Lean Formalization
14:30 - 15:00Coffee Break
15:00 - 16:30Mohammed Abouzaid, Yael Kalai, Jon Kleinberg, Raghu Meka
Panel Discussion
16:30 - 18:00Poster Session