Exclusive: GitHub CPO on multi-model Copilot and new AI tools

At its annual developer conference, GitHub Universe 2024, the platform announced several transformative updates, focusing on developer choice and more artificial intelligence (AI) integrations. Key highlights include the introduction of a multi-model Copilot system and the unveiling of GitHub Spark, a new tool designed to democratise app creation through natural language.

GitHub Copilot

During the conference, GitHub introduced its multi-model Copilot, which allows developers to select from models provided by industry giants like Anthropic, Google, and OpenAI. This update is a step towards giving developers more control over their workflows. "In 2024, we experienced a boom in high-quality large and small language models that each individually excel at different programming tasks," said GitHub CEO Thomas Dohmke. "There is no one model to rule every scenario, and developers expect the agency to build with the models that work best for them."

Mario Rodriguez, GitHub's Chief Product Officer, elaborated on how the choice of models can enhance development. "We have a default model overall, but providing developers with the option to select specific models for certain tasks fosters creativity and innovation," he said. GitHub currently employs OpenAI's 3.5 Turbo model for code completion, as it offers superior speed over newer versions. Rodriguez explained, "Depending on the task, we have specific requirements that guide us in selecting the optimal model. We're also expanding partnerships with companies like Anthropic and Google to offer even more choice in the coming weeks."

This multi-model approach aligns with GitHub's mission to empower developers to tailor their workflows, ensuring they use the most effective tools for each project. "Today, we deliver just that," said Dohmke. Anthropic's Claude 3.5 Sonnet model, which is known for its robust handling of complex coding challenges, was also integrated into Copilot. Jared Kaplan, co-founder of Anthropic, noted, "Through GitHub Copilot, Claude will help even more developers throughout the entire development process, from concept to deployment."

Enabling 1 billion developers

GitHub's announcement of Spark reflects its ambition to reach a broader audience of developers—both professionals and hobbyists. Spark allows users to create micro-apps, known as "sparks," using natural language prompts without needing extensive coding expertise. "Spark is not about low-code or enterprise solutions," Rodriguez clarified. "It's about enabling individuals to use natural language as a programming tool, whether they are experienced developers or just starting."

Rodriguez highlighted the personal impact of Spark: "I created an app with my daughters using Spark, and it showed how easy it is to turn ideas into reality without waiting for professional developers." Sparks can integrate AI capabilities and external data, making them versatile tools for personal or professional use. "This tool brings software development to the masses. With natural language as the interface, anyone can build an app on their phone or desktop and use it immediately," Rodriguez added.

GitHub envisions that this tool will help create "1 billion developers" worldwide by removing traditional barriers to entry. "Natural language is part of our DNA, and tools like Spark allow more people to unlock their creativity," Rodriguez said.

AI-native development

GitHub also announced several enhancements across its platform to support AI-native development. These include improvements to Copilot in Visual Studio Code, the introduction of multi-file editing capabilities, and new features in Copilot Workspace. Developers can now personalise Copilot responses with custom instructions and receive rapid code feedback within 30 seconds. Additionally, the expanded Copilot Extensions will allow users to interact with external developer tools such as Atlassian, Docker, and Stack Overflow.

"Working closely with developers, we've rolled out over 100 changes to Copilot Workspace," Rodriguez explained. "The improvements include brainstorming modes, deeper AI assistance, and better integration with tools like VS Code." GitHub's new Copilot Autofix tool, designed to automate security patches, was also showcased. "Developers are now fixing vulnerabilities three times faster with Copilot Autofix," Rodriguez noted.

The Octoverse 2024 report, released during the conference, highlighted the growing influence of AI on software development. Python has now overtaken JavaScript as the most popular language on GitHub, a shift driven by the rise of AI projects. "The language of AI has overtaken the language of the web," Rodriguez remarked. "Python's growth reflects the surge in global generative AI activity, particularly in regions like India, Germany, and Singapore."

Security and trust

Security and responsible AI usage were also key themes at the conference. Rodriguez stressed that GitHub prioritises transparency and privacy. "We guarantee our business users that their code won't be used without their consent," he said. "We've set up a Copilot Trust Center to provide developers with detailed information about our practices."

GitHub's commitment to security extends to all its tools, including Spark. "Like any tool, there's a learning curve, but we apply the same rigour to Spark as we do to our professional tools," Rodriguez assured. "The open-source community has embraced these efforts, and it shows the trust and love they have for GitHub."

Future vision

Rodriguez emphasised the importance of continuous learning and curiosity in software development. "Prompt engineering is becoming a vital skill as natural language interfaces gain prominence," he said. "Curiosity and the ability to remix existing tools and ideas will be essential for the next generation of developers."

GitHub's focus on fostering creativity and enabling innovation aligns with its broader mission to make software development accessible to all. "We want people to have fun with software, whether they're building something ephemeral with their family or solving complex problems at work," Rodriguez said.

"There's no blueprint for the future, but through constant partnerships and innovation, we can build something extraordinary," Rodriguez concluded.