Skip to Content.
Sympa Menu

lfcs-interest - Joint AIAI/LFCS Seminar- TODAY Monday 29th September 2025 by Visiting Speaker: Andrew D. Gordon

Subject: Events and news relevant to the Laboratory for Foundations of Computer Science

List archive

Chronological Thread  
  • From: Jodie Cameron <Jodie.Cameron@ed.ac.uk>
  • To: "inf-seminars@mlist.is.ed.ac.uk" <inf-seminars@mlist.is.ed.ac.uk>, "lfcs-interest@mlist.is.ed.ac.uk" <lfcs-interest@mlist.is.ed.ac.uk>, "AIAI Members" <aiai-members@inf.ed.ac.uk>, "ipab-members@inf.ed.ac.uk" <ipab-members@inf.ed.ac.uk>, "inf-ilcc-hcrc-seminar@mlist.is.ed.ac.uk" <inf-ilcc-hcrc-seminar@mlist.is.ed.ac.uk>, "ilcc-members@inf.ed.ac.uk" <ilcc-members@inf.ed.ac.uk>
  • Subject: Joint AIAI/LFCS Seminar- TODAY Monday 29th September 2025 by Visiting Speaker: Andrew D. Gordon
  • Date: Mon, 29 Sep 2025 08:50:58 +0000

 

Hello all,

 

There will be an joint AIAI and LFCS Seminar on Monday 29th September at 11am in Informatics Forum G.03.

 

Title: Vibe Checking: From Prompt to Protection

 

Speaker: Andrew D. Gordon, Cogna and University of Edinburgh

 

Abstract: Generative AI now enables both professional developers and domain experts to create software directly from natural-language requirements. Yet such software can still fail—either because the requirements themselves are incomplete or ambiguous, or because their translation into code is flawed. If “vibe coding” is prompting an AI to generate code, we advocate “vibe checking”, that is, prompting the AI to also generate code assertions. Our hypothesis is that systematic vibe checking helps guard against both specification errors and misinterpretation errors. Assertions—including preconditions, postconditions, and loop or data invariants—can be checked dynamically at runtime or verified statically, catching problems early. They can also be explained in plain language, enabling end-users to understand and correct the requirements themselves. This talk will survey the state of the art, share some current work at Cogna—an AI startup building precision software for physical industries with generative AI—and outline open research challenges in making AI-generated software truly trustworthy.

 

 

 

Bio: Andy Gordon is a computer scientist specializing in programming languages, AI, andhuman-computer interaction. Andy is Science Advisor at London startup Cogna where he leadsresearch on delivering software from natural language. Before joining Cogna as an early employee in 2023, Andy had a 26 year career at Microsoft Research. As Partner Research Manager, Andy led a diverse team of researchers and engineers to evolve Excel as an end-user programming language. He made significant contributions to the development of natural language formulas using generative AI in Copilot for Excel, formula features like LET/LAMBDA, the Calc.ts client-side execution engine for Excel formulas, and Excel Labs. Andy was recognised as a 2020 Fellow of the Association for Computing Machinery (ACM) for his research on programming languages: their principles, logic, usability, and trustworthiness. Andy is now Honorary Professor at the University of Edinburgh, following 12 years as full Professor. His PhD research at Cambridge contributed to the design of monadic I/O in Haskell, with his ASCII art “>>=" inspiring the Haskell logo.

 

Best wishes

Jodie

 

Jodie Cameron

Institute Administrator

School of Informatics

 

The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. Is e buidheann carthannais a th’ ann an Oilthigh Dhùn Èideann, clàraichte an Alba, àireamh clàraidh SC005336.


  • Joint AIAI/LFCS Seminar- TODAY Monday 29th September 2025 by Visiting Speaker: Andrew D. Gordon, Jodie Cameron, 09/29/2025

Archive powered by MHonArc 2.6.24.

Top of Page