The rise of AI-powered coding tools such as GitHub Copilot has sparked a familiar anxiety. If AI can write, debug, and even ship code, where does that leave the human developer?
At GitHub Constellation 2026, Jay Parikh, EVP of Microsoft, in an exclusive interaction with AIM, pushed back on that fear, saying Copilot is not replacing developers but changing how they build, collaborate, and think about software.
India now has 27 million developers on GitHub, with more than two million joining in 2026.
“We’re seeing tremendous engagement and growth from the community,” Parikh says. “India already has one of the largest developer communities in the world, and its level of open-source contribution is prolific.”
Developers in India have made more than 7.5 million contributions to open-source AI projects on the platform, making the country the second-largest contributor globally.
Parikh says the role of the developer, and even their day-to-day work, is evolving. “I don’t think it’s going away by any means. What we all need to do is stay curious, learn these technologies because these are not static,” he adds.
Should Developers Be Worried?
Parikh emphasises that while AI is changing how software is built, the fundamentals of engineering remain unchanged. Even if developers increasingly write specifications in natural language, those instructions must still translate into real systems running across databases, file systems, and networks.
At the same time, he notes that what has dramatically changed is speed. “The thing that’s changed is your ability to take a concept into production. That time has effectively collapsed,” he says. “But your knowledge of seeing that working system go through that phase is still necessary.”
As AI agents take on longer-running tasks, he believes developers will spend more time ideating, reviewing, and iterating together. “We could go write a specification, give it to an agent… and then it would be doing something for four hours or for the rest of the day,” explains Parikh, who worked at Facebook as VP and Global Head of Engineering for over a decade.
Beyond technical depth, Parikh stresses collaboration as a core skill. “Being able to ideate, create, and troubleshoot with colleagues, those communication skills are super important,” he says.
The executive adds that developers also need to become fluent in working with AI tools, whether it is writing effective prompts, building capabilities such as MCP servers, understanding how different models behave, or fine-tuning them. Skills in model adaptation and techniques such as reinforcement learning will be key to making these technologies truly useful in real-world applications.
The Rise of Agentic Software
Meanwhile, GitHub has moved from assisting with small tasks to handling far more complex, autonomous workflows. Earlier versions of Copilot focused on short, well-defined actions such as fixing bugs or refactoring code.
“Last year, you could give it a task, and it could go off and do something, but these were short-duration tasks like fixing a bug or refactoring a method,” he says.
That is now changing. Developers can now orchestrate multiple agents working on larger, long-running problems with richer context.
“In the last couple of months, you can actually manage and orchestrate a team of agents, and give them longer specifications and context,” he says. “They may go off and be working for hours or days on tasks.”
Notably, GitHub has been busy with agentic AI releases.
The company has expanded its agent ecosystem with Agent HQ, a centralised control layer that allows developers to deploy, manage, and switch between multiple agents, including those from external providers like OpenAI and Anthropic, within the same workflow.
GitHub also launched the Copilot SDK in technical preview, allowing developers to build and embed their own agents into applications and workflows. “I’m pretty fascinated by that because I think there’s just a whole category of experiences and applications that have yet to be built,” Parikh expresses.
The company also rolled out Copilot CLI, bringing agentic capabilities to the terminal. The tool introduces features such as plan mode, autopilot execution, and dynamic delegation, effectively turning the command line into an agent-driven development environment.
OpenClaw is a Game-Changer
Parikh describes his experience with OpenClaw as both insightful and reflective of how human interaction with AI is changing. “It was very eye-opening and very fun.”
The open-source and locally hosted AI agent can automate complex, multi-step tasks by controlling the computer’s browser, files, and applications.
Parikh says the tool points to a new phase beyond chat and reasoning models, moving toward more persistent and personalised agent experiences.
“It can be in WhatsApp and can do various things,” he says, adding that it can be personalised and that users can have fun with it, while it can also do real work depending on how it is configured.
He confirms that Microsoft is also building something similar.
The Big Tech company has already introduced related tools such as Copilot Cowork and Copilot Tasks, which automate workflows such as email management, scheduling, and other office functions.
Cowork allows users to assign complex and long-running tasks to AI agents.“You can give this agent harder, deeper and longer-running projects that it can go off and interface,” he adds.
Open vs Closed Models
Lately, the US has been slow to release open-source models, with many of the best ones coming from China. Models such as Qwen 3.5, GLM-5, and Kimi K2.5 have gained traction globally, with a focus on reasoning, coding, and agentic capabilities.
Parikh frames the open versus closed model debate as less of a competition and more of a question of choice for developers and enterprises. He says Microsoft Foundry brings together more than 11,000 options, from small to large models, across both open and closed ecosystems.
More jobs?
The rise of AI agents has sparked concerns about the shrinking number of developer jobs. Parikh sees a more nuanced future.
“There will be companies that try to do more with less,” he says. “But there will also be companies that understand the potential of AI and continue to hire aggressively.”
In fact, AI could unlock more entrepreneurship. “People who couldn’t build before can now create products and start companies,” he says. “That’s going to increase innovation.”
Parikh reiterates that the developer’s role will not disappear.
“There’s never been a better time to be a builder,” he says. “The tools are accelerating us, but it’s still up to us to imagine and create what comes next.”
- OpenAI Says AGI is Here. But Not Everyone is Buying It
- Why Oracle is Bringing Post-Quantum Security to Java
- What Really Happens When You Turn Your Selfie Into a 1980s AI Pic?
- AI Agents Are Making Data Security A Bigger Enterprise Problem
- OpenAI is Getting Nervous About Reinforcement Learning
- Meta’s Bold Open-Weight Bet is Meant to Muse Developers