At WeAreDevelopers in Berlin, Thomas Dohmke delivered a compelling keynote on how AI is reshaping software development. Drawing on Microsoft’s decade‑long journey in developer tools and GitHub’s rapid innovations, he showcased Copilot’s transformation from a simple autocomplete feature to a sophisticated coding agent capable of streamlining the entire development lifecycle.
When Microsoft acquired HockeyApp in December 2014, Dohmke moved to Seattle the following summer, bringing with him a deep background in developer tools. More recently, GitHub’s purchase in 2018 marked a pivotal moment for the company’s commitment to AI-driven development. According to Dohmke, these milestones underscore Microsoft’s enduring mission: empowering developers at every stage of their work.
Beginning life in 2020 as an autocomplete tool, Copilot quickly captivated users with its ability to predict code snippets. By 2022—coinciding with Dohmke’s inaugural appearance as GitHub CEO—the tool was responsible for generating 40 percent of code in enabled files. The release of ChatGPT in November 2022 accelerated this progress: Copilot Chat, CLI integration, and voice support arrived soon after, enabling natural language prompts in languages ranging from German and French to Hindi and Bahasa. Since then, the introduction of multi-modal choices, multi-file edits, and Model Context Protocol (MCP) support has elevated Copilot into a full-fledged AI agent, capable of tackling complex, asynchronous tasks across projects.
During the keynote, a live demo illustrated how assigning “migrating tests” to the Coding Agent triggers an automated workflow: a virtual machine spins up, the repository is cloned, analysis runs, and a pull request draft appears—all without blocking the developer’s work. Another demonstration featured Agent Mode within the IDE, where the AI autonomously navigated the codebase to add a snake game feature, executing tests and iterating based on feedback. Throughout these examples, Dohmke emphasized that final approvals and code validations remain firmly in human hands, ensuring quality and fostering a true partnership between developers and AI.
Survey data revealed that Copilot feels like a “superpower” to its users: 52 percent reported coding faster than ever—“faster than a speeding bullet”—while others likened it to having “X‑ray vision” or the ability to “leap a building in a single bound.” Beyond sheer speed, Copilot’s multilingual support is breaking down barriers, allowing even young learners to code confidently in their native tongue. Feedback describing the experience as “mind‑blowing” and “truly amazing” reinforces a core message: AI augments human creativity rather than replacing it.
Technical hiccups during live demos—such as an HDMI failure in 2022 and errors in the snake game showcase—served as reminders that AI agents still require human oversight. Developers must remain ready to debug, validate, and guide agents, preserving the essential skills that underpin reliable software creation. Persistent skepticism among some practitioners underscores the need for continued education and hands‑on experience with these emerging tools.
Looking ahead, Dohmke painted a picture of hybrid teams where humans collaborate seamlessly with specialized AI agents—whether for coding, code review, or security auditing. This vision aligns with GitHub’s ambitious goal to empower “1 billion developers,” making software creation accessible to anyone, not just seasoned engineers.
Dohmke closed by declaring that “the world has won if we generate happy developers,” a fitting capstone to Copilot’s journey from simple autocomplete to a full‑blown AI agent that accelerates coding, embraces multilingual workflows, and forges seamless human‑AI collaboration.
WeAreDevelopers, held annually in Berlin, convenes thousands of software engineers, tech leads and industry innovators to explore the latest in development and emerging technologies. Our team was on the ground in 2025, and we’ll be sharing more insights on our blog and LinkedIn in the days ahead.