Tamás Szabó

Exploring how to effectively partner with AI to accelerate software development - A tour of recent GitHub Next experiments

14.00-14.50

While AI models have immense knowledge in general, it seems that effectively harvesting that to aid software developers is a challenging task. Different prompting techniques, tailored user interfaces, and various modalities all come into play when we want to create a developer tool that is actually useful. Our goal, at GitHub Next, is to conduct experiments and create technical previews to understand the solution space around AI-assisted developer tools. This talk is about some of our recent experiments. First, I talk about GitHub Spark, which is an AI-powered tool for creating and sharing micro apps, which can be tailored to your exact needs and preferences, and are directly usable from your desktop and mobile devices. Spark does not require you to write or deploy any code. The second part is about Copilot Workspace, which is a task-oriented development environment powered by AI. I present briefly the evolution of the design of Copilot Workspace that was governed by all the user feedback we got since the technical preview phase of the tool has started. Finally, I will dive deeper into some of the agentic interactions that drive key building blocks of Copilot Workspace; how we use vision models to understand images in the task description, or how we converse with the model to identify the slice of a code base that is relevant to solving a task.

Tamás works as a research engineer at GitHub Next, which is a research lab within GitHub exploring the future of software engineering. He is involved in a number of Next explorations centered around (AI-assisted) developer tooling. He holds a PhD in computer science, and his research topic was about the efficient execution of program analyses. His work has been presented in scientific publications, and you can regularly find him speaking at industry events, as well. Outside of work, you can either find him spending time with his wife and son or riding his mountain bike.