GitHub COO Kyle Daigle Explains How AI Is Changing Coding

Artificial Intelligence is changing how software is built, lowering the barriers to programming and enabling more people to turn their ideas into applications. As AI coding agents take on tasks such as writing code, testing applications and documenting projects, developers are increasingly shifting their attention towards defining problems, managing systems and ensuring that software works as intended.

Microsoft-owned GitHub is at the centre of this transformation. The code repository platform has seen growing adoption of AI-powered development tools, including GitHub Copilot, as experienced programmers and newcomers experiment with automated coding.

In an interview with The Indian Express on the sidelines of GitHub Universe 2026 in San Francisco, GitHub Chief Operating Officer and Chief Marketing Officer Kyle Daigle discussed how AI agents are changing Software Development, the skills future developers will need and the opportunities emerging for people without traditional computer science backgrounds.

Daigle said GitHub is seeing more developers build and publish projects than ever before. He also emphasised that generating code faster does not automatically translate into better software, making verification, quality assurance and problem-solving increasingly important.

GitHub Reports Growth in Developers and AI Coding Adoption

According to Daigle, GitHub’s developer community and the use of its AI-powered tools have continued to expand. He said the platform had reached 225 million users, including 50 million people using GitHub Copilot every day.

Daigle also reported that developers made 13.5 billion contributions on GitHub in 2026, more than the combined total for 2024 and 2025. The figures reflect the scale of activity on the platform as AI tools become increasingly integrated into software development workflows.

He said the growth extends beyond developers working at large companies. Open-source maintainers are also using AI tools to build projects and contribute to existing software, expanding the range of people participating in software development.

GitHub, he explained, has traditionally served developers regardless of the tools they use or the applications they build. AI agents are now extending that role by making it easier to start new projects, modify existing applications and add functionality without manually writing every line of code.

Daigle suggested that the growing availability of these tools is encouraging people taking their first steps into programming to begin building on GitHub. The change could broaden participation in software development by allowing people with different levels of technical experience to experiment with ideas that previously required more specialised coding knowledge.

AI Agents Are Changing the Developer’s Role, Not Eliminating It

AI coding agents can increasingly perform multiple stages of development, including generating code, writing documentation, running tests and assisting with technical decisions. This has prompted questions about whether developers will continue to need the same level of hands-on programming experience.

Daigle compared the current transition to earlier changes in programming technology. Developers have moved from punch cards to compiled languages, dynamic languages and increasingly sophisticated development tools. Each change has altered how people write software and the level of abstraction at which they work.

He argued that these developments have allowed programmers to concentrate on more complex problems rather than spending all their time writing code manually. AI agents represent another stage in that progression, potentially allowing developers to accomplish more than would be possible through traditional coding methods alone.

However, Daigle stressed that automation does not remove the need to understand the problem being solved or how the resulting software should function.

Developers still need to determine what they want to build, understand how its components fit together and assess whether the finished application meets the intended requirements. They must also consider whether the software can operate reliably at the scale for which it is designed.

In this environment, the developer’s role increasingly involves directing automated systems, making informed technical decisions and evaluating the results. AI can help implement a solution, but human understanding remains important when deciding whether that solution is appropriate and reliable.

Why Faster Code Generation Does Not Guarantee Faster Software Delivery

One of the central points in Daigle’s assessment is that producing code has never been the only challenge in software development. Deciding what needs to be built and determining whether it has been built correctly can be equally difficult.

AI agents can generate applications quickly, but the amount of code produced is not necessarily a measure of productivity. If developers cannot establish whether the output works correctly, meets user requirements or contains security vulnerabilities, faster generation may simply create more work later in the development process.

Daigle said GitHub is focusing on helping developers manage the entire journey, from planning and defining requirements to verifying the finished product.

He illustrated the point with an example of asking GitHub Copilot to create an application that ranks the emojis a user employs. Generating the application is only the beginning. A more complete workflow would allow the agent to inspect the application, interact with its interface, deploy it, confirm that it runs correctly and scan it for security vulnerabilities.

If the system identifies security issues, it should also help address them. This approach moves AI-assisted development beyond writing code towards checking whether the resulting application is usable, secure and consistent with the original objective.

Daigle said GitHub’s code-quality tools can automatically remediate certain issues with Copilot. He also pointed to Copilot Automations, which allows developers and organisations to create customised workflows in response to events on the platform.

For example, a team can establish verification processes triggered when a pull request is opened, an issue is closed or another platform event occurs. These workflows can help ensure that appropriate checks take place as software changes.

The broader objective is to make verification an integral part of development. Rather than concentrating exclusively on producing more code, developers can use automated checks to reduce the time spent identifying and correcting problems.

GitHub Copilot Usage Expands as Multiple AI Agents Work Together

Daigle said GitHub Copilot had continued to grow since the introduction of its AI agent capabilities. Alongside the 50 million users he cited, he reported adoption by 140,000 organisations worldwide.

He also said usage of Copilot’s cloud agent had increased by 230 per cent. These figures, according to Daigle, reflect the growing role of AI agents in development workflows.

One important change over the past year has been the move away from a simple one-to-one interaction between a developer and an AI assistant. Previously, a developer might assign a task to an agent, wait for the result and then decide what to do next.

GitHub’s approach is increasingly built around running multiple coding agents simultaneously. Daigle said developers can work with these agents through several interfaces, including the GitHub Copilot app, Visual Studio Code, the Copilot command-line interface and GitHub.com.

In some cases, multiple agents can operate within a single project. This allows developers to delegate different tasks and manage more complex workflows, although the results still need to be checked for correctness and consistency.

Daigle said the most important lesson from the past year was that GitHub should not focus simply on helping developers generate more lines of code. The emphasis must be on ensuring that generated code is necessary, works as intended and can be verified.

This reflects a broader change in how productivity is measured. Reusing libraries and existing components has long helped developers avoid unnecessary work. AI tools can expand those efficiencies, but only if the code they generate contributes to a working and maintainable application.

Will AI Make Programming Languages Such as Python Obsolete?

The growing role of AI coding agents has also raised questions about whether developers will still need to learn programming languages such as Python, JavaScript and TypeScript.

Daigle does not expect these languages to disappear. Instead, he believes AI agents will increasingly influence which languages developers choose for particular projects.

Programming languages offer different technical characteristics, including differences in performance, memory usage and deployment requirements. As AI agents become more capable of working across languages, developers may be able to select a language according to a project’s needs rather than relying primarily on the languages they already know.

Daigle cited the rewriting of GitHub’s Copilot CLI from TypeScript to Rust as an example of how AI assistance can support changes to a project’s technical foundation.

He expects further shifts in language choices over the coming years. However, his comments do not suggest that understanding programming fundamentals will become unnecessary. Instead, the ability to evaluate technical requirements and make sound decisions may become more important as agents handle a greater share of implementation work.

For aspiring developers, this means that learning a language remains useful, but memorising syntax alone is unlikely to be enough. Understanding how software works, how different components interact and how to identify problems provides a foundation for using AI tools effectively.

Problem-Solving and Systems Thinking Will Remain Essential Skills

Daigle identified three important areas for people entering software development: understanding programming fundamentals, systems thinking and problem-solving.

He said beginners should start by identifying a problem and breaking it into smaller, manageable parts. This approach helps them understand what a solution needs to accomplish before deciding which tools or programming languages to use.

The same principle applies to experienced developers adopting AI agents. Although an agent may generate an implementation, someone still needs to establish the requirements, understand the wider system and determine whether the result solves the original problem.

Systems thinking is particularly important when applications involve multiple components, external services or complex workflows. A change that appears correct in isolation may create problems elsewhere, making it necessary to understand how the different parts of a system interact.

Problem-solving also helps developers investigate why an application behaves unexpectedly. Asking why a particular implementation was chosen, why a deployment failed or why software does not work on a specific computer can turn an AI-generated result into an opportunity to develop deeper technical understanding.

Daigle said these fundamentals are relevant to people at different stages of their careers, including experienced programmers, beginners experimenting with coding agents and those hoping to become software developers.

As AI tools become more capable, the ability to describe a problem clearly, evaluate possible solutions and verify the outcome can help users make better use of automation without relying on it blindly.

Software Development Could Become More Focused on Verification

Looking ahead, Daigle expects software development to place greater emphasis on giving AI models enough context and capability to verify that applications work as intended.

Traditional development already relies on unit tests and automated testing to check software behaviour. Daigle expects these methods to remain important, while AI agents increasingly help test whether an application meets the user’s actual requirements.

This distinction matters because passing individual tests does not necessarily mean that an application fulfils its intended purpose. Software may perform correctly under specific test conditions but still fail to deliver the experience or functionality users expect.

AI agents could help address this gap by testing applications in ways that more closely reflect their intended use. If an agent can understand the original requirements, interact with the application and assess the results, it may help developers identify problems earlier in the process.

However, the effectiveness of such systems depends on the quality of the context they receive and the reliability of their verification methods. The goal is not simply to automate more steps, but to establish whether the final product is correct and useful.

Daigle expects this approach to accelerate development by helping teams build applications that are more likely to work as intended, rather than merely increasing the volume of code produced.

AI Coding Tools Could Bring Non-Technical Professionals Into Development

One of the most significant changes Daigle anticipates is the growing participation of people who do not consider themselves software developers. Over the coming year, he expects more knowledge workers and information workers to build applications for personal use, family members or friends.

These projects could include mobile applications and web apps designed to solve everyday problems. With AI coding assistants, people may be able to turn an idea into a working prototype without first mastering every technical detail of implementation.

Daigle believes this could significantly reduce the barrier between people who can program and those who have traditionally viewed coding as beyond their reach. He described this as a way to democratise software creation, particularly for personal computing projects.

That does not mean every beginner will immediately be able to build complex, production-ready systems without technical knowledge. Understanding requirements, checking results and dealing with unexpected behaviour remain important. The change is that people can begin experimenting and learning without needing to acquire all their programming skills in advance.

AI agents can also make the learning process more interactive. Instead of relying entirely on books or formal courses, beginners can start with a specific problem, ask questions while building a solution and investigate why different approaches work or fail.

This could be particularly valuable for people whose professional backgrounds are outside technology. Experience in design, construction, languages, the arts or other fields may help them identify problems worth solving and develop applications suited to their areas of expertise.

GitHub COO Says a Computer Science Degree Is Not the Only Route Into Tech

Daigle drew on his own experience to explain why a traditional computer science degree is not necessarily a prerequisite for starting a career in software development. He said he joined GitHub 13 years ago without having a computer science degree.

Before joining the company, he had written code to address problems in his town, including voter registration, using Access databases. His experience led him to challenge the idea that aspiring developers must complete a formal computer science education before they can begin building software.

Daigle recalled buying lengthy books that attempted to teach programming languages such as C from beginning to end. At the time, he believed that finishing such a book would make him a developer. He now sees learning to code as only one part of the process.

For beginners, he said, AI coding tools provide a more accessible starting point. Users can create a GitHub account, begin experimenting with Copilot and start by answering a simple question: what do they want to build?

From there, they can define the problem, describe how they want the solution to work and use the tools to help implement it. Along the way, they can ask questions about technical choices, deployment and errors.

Daigle encouraged people to draw on their existing interests and experience while learning. Curiosity and a willingness to investigate unfamiliar concepts can help beginners build practical skills even if they did not follow a conventional technical education path.

He also stressed that companies are looking for people who are curious, passionate and willing to learn as the industry changes. AI tools may make it easier to begin developing software, but continued learning remains important for people who want to turn their experiments into professional opportunities, businesses or startups.

India Emerges as a Major Developer Community for GitHub

India is an important part of GitHub’s expanding developer ecosystem. Daigle said the country added 9.34 million developers to the platform over the preceding year, taking its community to around 31 million developers.

He said developers in India are working across startups and enterprises, including on sovereign AI initiatives and increasingly ambitious software projects.

The growth reflects the expanding participation of Indian developers in a software industry being reshaped by AI. As coding tools become more accessible, they may also create opportunities for students, professionals from non-technical backgrounds and entrepreneurs who want to build applications around problems they understand.

Daigle said GitHub wants to make software development easier through tools such as the Copilot app and workflows that allow multiple AI agents to work on projects with broader context.

For India, the potential benefit extends beyond experienced software engineers. People who have domain knowledge in areas such as business, education, design or local services may be able to use AI tools to develop applications that address specific needs. Their existing expertise can help them identify the problem, while technical learning and verification help turn an idea into a functioning product.

Nevertheless, easier access to coding tools does not remove the need to evaluate software quality. Developers and organisations must still ensure that applications meet their requirements, behave reliably and address relevant security concerns.

Multiple AI Models Could Give Developers More Choice

Daigle also expects the growing availability of models from different AI laboratories to influence how developers use coding assistants. He pointed to the emergence of open-source, open-weight and sovereign models as part of a broader shift in the AI ecosystem.

He argued that relying on a single model or model provider may not be the best way for developers to access advanced AI capabilities while maintaining control and choice.

As more models become available, developers and organisations may have greater flexibility in selecting tools for different tasks. The choice of model can become another part of the software development process, alongside decisions about programming languages, architecture and deployment.

This fits with Daigle’s broader view that AI agents will increasingly handle implementation details while developers focus on the problems they need to solve, the systems they are building and the methods used to verify the results.

What AI’s Expansion Means for Future Developers

Daigle’s comments point to a shift in software development from manually writing every line of code towards coordinating AI tools, defining requirements and verifying completed work.

For experienced developers, coding agents may offer ways to handle repetitive tasks, explore alternative implementations and manage several parts of a project simultaneously. For beginners, the tools can provide an accessible route into programming by allowing them to start with practical problems rather than waiting until they have mastered a language.

But the underlying principles of software development remain relevant. Developers still need to understand what they are building, break problems into manageable steps, choose appropriate tools and determine whether the final application works correctly.

As AI-generated code becomes more common, verification and quality assurance could become even more central to professional development workflows. The ability to generate code quickly will be useful, but the ability to establish that the code is necessary, secure and fit for purpose will remain crucial.

For people in India and elsewhere who are considering a career in technology, the changing landscape offers new ways to begin learning and experimenting. A traditional computer science background is not the only starting point, according to Daigle, but curiosity, problem-solving and a commitment to learning will continue to matter.

Disclosure: The writer attended GitHub Universe at the invitation of GitHub US. Microsoft covered travel and accommodation expenses.