Diego Kingston, CCSD's Research Director, presented the paper "Preliminary Projects with Real Clients in AI-Assisted Programming Workshops" at CADI CAEDI 2026 this week, co-authored with Pablo Deymonnaz, Leandro Ferrigno, and Juan Bono. The event, which brings together the 8th Argentine Congress of Engineering and the 14th Argentine Congress of Engineering Education, took place in San Fernando del Valle de Catamarca.
The paper describes a shift in approach in Programming Workshop (Taller de Programación), the course at the University of Buenos Aires School of Engineering that for years had every team implement the same assignment (RoomRTC, BitTorrent, Bitcoin) with no real client involved. This semester, that changed. Each team drew up a preliminary project in response to the concrete need of a faculty member or researcher from another discipline within the same school.
In the end, eight cases were completed. Six were for the Department of Surveying, among them a bathymetric simulator integrated with an augmented reality sandbox and a GNSS data converter compiled to WebAssembly. One was for Physical Chemistry: a suite of 20 laboratory simulators with a five-layer architecture. And the last was for Mechanical Engineering: a server that centralizes control of nine 3D printers using Klipper and Moonraker. Everything was developed in Rust, under the same quality constraints the course has always required, and with generative AI organized into four stages: domain research, design, implementation, and validation with the client.
The goal is for Computer Engineering students to negotiate scope with someone who does not think in technical terms, instead of delivering a system that only the teaching staff will evaluate.