What's Next for GitHub?

Key Concepts

• GitHub Universe: GitHub's annual developer conference.
• GitHub Copilot: An AI-powered code completion tool.
• Agents (AI Agents): Autonomous AI programs designed to perform tasks.
• SDLC (Software Development Life Cycle): The entire process of developing software, from planning to maintenance.
• Developer Choice: The ability for developers to choose their preferred tools and workflows.
• Asynchronous Work: Collaboration and communication that doesn't require participants to be online simultaneously.
• CodeQL: A semantic code analysis engine.
• GitHub Actions: A CI/CD platform that allows automation of software development workflows.

GitHub Headquarters Visit and Universe Conference

The video features a visit to GitHub headquarters with COO Kyle Dagel. The discussion centers around the upcoming GitHub Universe conference, which is expected to be the largest yet, hosting 3,700 developers. Dagel describes the venue at Fort Mason as a modernized pier, maintaining a "builder theme" reminiscent of past, more industrial locations.

Changes at GitHub and Continued Focus on Developers

A significant portion of the conversation addresses recent changes at GitHub, particularly following the departure of Thomas Dohmke. Dagel emphasizes that GitHub's core mission of focusing on developers remains unchanged. He states that the company has, in fact, received increased resources, both figurative and literal, to fulfill this mission, citing GitHub Copilot as an example. He highlights the close partnership with the VS Code team and the conviction that GitHub's unique intertwined business model across enterprise and open-source developers is a strength. Dagel reassures employees and the community that the fundamental principles and support for developers are consistent, with an increasing amount of support to keep developers at the forefront.

Infrastructure Migration to Azure and Developer Concerns

Concerns are raised about the migration of GitHub's infrastructure to Azure. Dagel acknowledges these concerns, drawing a parallel to the skepticism following the Microsoft acquisition. He explains that the move to Azure is driven by the need for more compute power due to rapid growth and that it makes strategic sense. He reiterates that GitHub will continue to choose tools that provide an advantage to developers, even if they are outside the Microsoft ecosystem. The goal is to prove, through continued focus on developer tools and ethos, that GitHub remains committed to its nearly 20-year-old open-source and developer-centric approach.

Kyle Dagel's Role and Philosophy

Dagel discusses his evolving role at GitHub, moving from engineering leadership to M&A, strategy, and communications. He expresses his passion for staying connected to the "steering wheel" of what makes GitHub unique and his commitment to listening to developers, even their concerns. He believes GitHub's success is rooted in its "humanity" – people who genuinely care about the mission beyond just shipping products. He sees his role as driving meaningful change that simplifies developers' lives and preserves GitHub's "authentic spark."

GitHub's Centrality and Responsibility

Dagel describes GitHub as being at the "center of a lot," likening its downtime to a "snow day" for the tech world, highlighting its critical role. He emphasizes the enormous responsibility of being a platform for "builders" and the need to balance this with the growing number of software developers and aspiring coders.

The Democratization of Code and the AI Era

Dagel shares his personal journey into coding, starting at age 12, and how access to a computer and the inspiration from his mother's work with rocket scientists made it seem attainable. He expresses excitement about the AI era, particularly how it's making coding more accessible. He notes that tools like "vibe coding" (likely referring to AI-assisted coding without deep technical knowledge) allow people to get started and build things they envision. He cites an example of a non-developer building conference dashboards using Copilot, demonstrating the power of AI in lowering the barrier to entry. This is in addition to enabling professional developers to build, operate, evolve, and scale software.

Balancing Professional Development and AI Tools

The conversation delves into balancing the needs of professional developers seeking advanced tools with the rise of AI-powered tools like Copilot and "vibe coding." Dagel states that GitHub's core mission is to serve anyone interested in code. He emphasizes that even with AI generating most of the code, understanding the underlying code is crucial for scalability. He discusses the need to mesh existing technologies with new AI approaches to create easier pathways for developers. He uses CodeQL as an example of a powerful tool that could be integrated with AI to offer broader capabilities beyond security analysis. The process involves testing, experimenting, and then locking in features for developers.

Integrations and Developer Choice

Dagel addresses the increasing number of companies, including competitors like Anthropic and OpenAI, building on top of GitHub and integrating tools into its platform. He views this as a positive development, stating that more tools lead to better outcomes for developers. He reiterates the "home for all developers" philosophy, which necessitates inviting everyone in. While acknowledging the strategic risks of allowing developer choice, he believes it drives innovation and ensures that developers choose GitHub's tools because they love them, not because they are mandated. He emphasizes that GitHub provides APIs and allows developers to use their data and preferred tools (observability, CI, etc.), reinforcing the commitment to developer-first principles.

GitHub Copilot's Evolution and Competitive Landscape

Regarding the competitive landscape for generative AI coding tools, Dagel acknowledges that while mindshare might be with other tools, usage is key. He explains that Copilot initially focused on autocomplete as a "pair programmer." As other models emerged, they specialized in different areas. GitHub has been experimenting with Copilot across various surfaces (IDE, CLI) for the past 18-24 months, focusing on providing full context and leveraging multiple models. The goal is to make Copilot valuable across the entire SDLC, not just code generation, recognizing that developers use a mix of tools. He highlights the potential for Copilot to leverage a developer's entire GitHub history (personal repos, pull requests, comments) to provide a truly personalized experience, moving beyond isolated tool functionalities.

Developer Demands and the Future of GitHub

Dagel outlines two broad categories of developer requests:

1. Core GitHub Evolution: Continued investment in fundamental features like pull requests, issues, repository interactions, and project discovery, always prioritizing the developer.
2. AI "Magic Moment": Developers are looking for AI tools that provide a "magic moment" of success, not just code generation but also intelligent assistance in problem-solving and identifying overlooked considerations. He stresses that "garbage in, garbage out" still applies, and AI should help developers with the "what to build" and "problem to solve" aspects, not just the conversion to code.

Addressing Dichotomies in AI Tool Adoption

Dagel addresses the dichotomy between developers who find AI tools highly effective and those who feel they don't work. He attributes this to individual experiences and the level of investment in learning the tools. He explains that users who invest time in prompt engineering, custom instructions, and understanding tool intricacies tend to have better experiences. He aims to build products that don't require such steep learning curves, comparing it to the evolution from C to Python for accessibility. The goal is to make AI tools more intuitive, allowing developers to get value without needing to be prompt engineering experts.

The Role of Agents and Asynchronous Work

The discussion shifts to agents and their impact on GitHub and Copilot. Dagel notes that agents are more productive in coding due to the verifiable nature of code. He emphasizes the importance of the human factor in the loop and the need for verifiable correctness, distinguishing agents from general chatbots. He sees agents moving beyond simple task execution to a more collaborative role, where developers can compare and contrast agent outputs. The focus is on managing the "abundance" of agent capabilities, enabling review, tracking, and steering of agents without constant monitoring.

Dagel highlights the importance of asynchronous work for agents, drawing a parallel to GitHub's own remote and asynchronous culture. He believes agents should enhance this asynchronous workflow rather than being tied to constant real-time monitoring. He envisions agents extending beyond coding to all aspects of GitHub's internal operations, including communications, marketing, and legal. He mentions features like Copilot Chat, custom instructions, and the potential for GitHub issues to track meeting minutes and be queried by Copilot. The goal is for agents to have full context and awareness across all GitHub platforms, enabling two-way engagement and automation, such as automatically closing issues when to-dos are complete.

GitHub Actions and Compute for Agents

The conversation touches upon GitHub Actions and the increasing need for compute power to run background processes and agents. Dagel explains that GitHub Actions are used by partners and for running parts of agents. He discusses the need for more options for custom images in Actions to support agents, especially for those with private dependencies. He also addresses the speed of agent execution, noting that while CI job launch times are acceptable, agent response times need to be incredibly fast. GitHub is investing in faster compute and containerization to enable quick agent workflows.

Monetization and Future of GitHub Universe

The monetization of these services is discussed, moving from seat-based to token-based models (per day/month). Dagel aims for straightforward yet flexible pricing, focusing on model choice and underlying technology. He mentions Copilot Pro and Pro Plus subscriptions. The goal is to make usage clear and simple, allowing users to focus on solving problems.

Looking ahead to GitHub Universe, Dagel anticipates a "new chapter" for GitHub, focusing on how agents will integrate into developer collaboration. He emphasizes building on the foundation of collaboration while enabling agents to work with developers rather than being separate entities. He reiterates his excitement about showcasing "developer choice" at Universe, aiming to surprise and delight developers with GitHub's commitment to this principle.

Conclusion

The interview with Kyle Dagel provides a comprehensive overview of GitHub's current state and future direction. Key takeaways include an unwavering commitment to developers, even amidst significant organizational and technological shifts. The integration of AI, particularly through GitHub Copilot and agents, is seen as a transformative force that will democratize coding and enhance developer productivity across the entire software development lifecycle. GitHub aims to be the central hub for this evolution, prioritizing developer choice and seamless integration of tools, while continuing to invest in its core platform and foster an asynchronous, collaborative environment. The upcoming GitHub Universe conference is positioned as a pivotal event to unveil further advancements in these areas.