Talks

Talk April 9

One Developer, Two Dozen Agents, Zero Alignment

Why we need collaborative AI engineering tools, and a tour of Ace — a research prototype from GitHub Next that brings teammates and coding agents into one shared, multiplayer workspace.

---

Talk October 27, 2023

Bringing the power of AI into your application

A practical introduction to integrating large language models into applications, covering prompt crafting, chaining AI queries, and UX considerations for building AI-powered tools. Also addresses how to measure the impact of AI changes and evaluate whether AI integration is right for a given application.

---

Talk May 11, 2023

Designing for Intelligence

A look at the design and research principles behind building AI-powered products at GitHub Next, covering how to elicit context from users, structure prompts, measure result quality, and weave intelligence into product experiences. Shares practical lessons learned from prototyping and shipping tools like GitHub Copilot across the software development lifecycle.

---

Talk January 12, 2023

GitHub Copilot and techniques for improving code suggestions from large language models

An examination of how to improve the performance of a given LLM on a specific task without modifying the underlying model — a more accessible area of research than training from scratch. Focuses on techniques used to enhance GitHub Copilot’s suggestions and introduces an offline evaluation framework for early-stage experimentation.

---

Talk November 21, 2022

A Backstage Pass with GitHub Copilot (Universe 2022)

An inside look at the research and data behind GitHub Copilot, combining live demos with survey results from GitHub engineers and academic perspectives on AI pair programming. Presents findings from a controlled study showing Copilot users completed a coding task significantly faster than those without it.

---

Talk October 27, 2022

AI-assisted pair programming with GitHub Copilot

An introduction to GitHub Copilot as an AI pair programming tool and how it integrates into developer workflows to accelerate coding. Covers how the system suggests code in context, adapts to individual projects, and what this means for the future of software development.

---

Talk October 25, 2022

Crafting Code Suggestions using Large Language Models

An exploration of why large language models for code completion present unique engineering challenges — developers don’t write code linearly, so models must bridge the gap between predictive capability and real-world usefulness. Covers prompt engineering strategies including content reordering and codebase linearization, using GitHub Copilot as a case study.

---

Talk September 10, 2022

How We Evaluate the Quality of Copilot

An overview of the methods and metrics used to assess the quality and effectiveness of GitHub Copilot’s code suggestions at scale. Covers how evaluation frameworks are designed to measure both objective output quality and developer productivity impact.

---

Talk June 16, 2022

Bringing the Power of AI into your Editor with GitHub Copilot

A technical deep dive into how GitHub Copilot works as an advanced AI autocomplete system embedded directly in the IDE, drawing on the collective knowledge of millions of developers. Covers the design principles behind Copilot, how it communicates with its AI model, and how it integrates with various code editors.

---

Talk June 13, 2022

Productivity Assessment of Neural Code Completion

A study examining how neural code completion tools affect developer productivity by comparing user perceptions with objective usage metrics across a real-world user base. Finds that suggestion acceptance rate — rather than whether completions persist in the codebase over time — is the strongest driver of developers’ perceived productivity gains.

---

Talk May 30, 2022

From AI for Code to AI for Coding

Large language models like OpenAI Codex have reached the capability to generate large chunks of coherent code, enabling a new class of developer tools including GitHub Copilot. Examines what work remains to transform a code-prediction model into a genuinely transformative programming assistant, and the obstacles along the way.

---

Talk April 7, 2022

Bringing data into your repos using Flat Data

Flat Data simplifies working with datasets in git by making it easy to fetch data from anywhere on the web directly into your repositories for processing, analyzing, and versioning. Builds on “git scraping” techniques to enable new developer workflows around local datasets.

---

Talk August 3, 2021

A Tour of F# Editor Tooling

A tour of the editor tooling ecosystem for F#, covering language server support, IDE integrations, and the developer experience across editors.

---