CSE Colloquium – Messages from across the event horizon:  AI Agentic Design for Computer Architecture (and more generalizable learnings)

Presenter: Christopher Fletcher, UC Berkeley

Abstract:

It is difficult to escape the hype of agentic coding.  Is the hype real?  Are we still living in ~Summer 2025 — when AI coding would accomplish little more than upset its human supervisor?  Or has a level shift in technology finally arrived?

In this talk I will argue the latter.  I will describe a self-imposed experiment to discover modern AI coding tools’ capabilities (starting mid February 2026).  I will try (my best) to communicate my utter and sheer surprise at where the state of the art actually is.  Then I will do a deep dive and try to relay everything I have learned about this new engineering discipline—based on my attempts to push the technology as hard as I can for the past 1.5 months.  I will conclude by pontificating about the future of computer architecture and academic research more generally.

Bio:

Christopher Fletcher is an Associate Professor of EECS at UC Berkeley. He is a computer architect whose research spans architecture, security, and domain-specific acceleration, especially at their intersections from cryptography and hardware attacks to algorithm-to-hardware co-design. His work has received 31 paper recognitions and several other honors, including the NSF CAREER Award, Intel and Google faculty awards, UIUC research and promotion awards, election to DARPA ISAT, and MIT’s George M. Sprowls Award, with related work also recognized by Scientific American as one of ten “World Changing Ideas.”

Hosted by: Professor Alvaro Cardenas

Location: Engineering 2, Room E2-180 (Refreshments such as fruit, pastries, coffee, and tea will be provided.)

Zoom Option: https://ucsc.zoom.us/j/93445911992?pwd=YkJ2TQtF79h0PcNXbEcpZLbpK0coiY.1&jst=3