Ekonomia Online
LATEST

“Writing code is over” – Microsoft engineer explains how AI is changing programming

“Writing code is over” – Microsoft engineer explains how AI is changing programming

“Writing code is over.” This bold statement by a Microsoft engineer has sparked discussion in the technology industry, but behind it lies a much bigger change: artificial intelligence is taking over more and more of the work that until now was done directly by programmers. David Fowler, Microsoft Distinguished […]

“Writing code is over.” This bold statement by a Microsoft engineer has sparked discussion in the technology industry, but behind it lies a much bigger change: artificial intelligence is taking over more and more of the work that until now was done directly by programmers.

David Fowler, a Microsoft Distinguished Engineer and one of the leaders of the Aspire project, summed up the change in a post on X, writing: “Typing code is absolutely over.”

But Fowler is not saying that programmers will disappear. On the contrary, the idea is that their role is changing. Instead of a developer spending hours writing code line by line, more and more of this work can be delegated to an artificial intelligence agent.

From suggestions to building an application One example of this is GitHub Copilot, which can now do much more than simply suggest code, Telegrafi reports.

Copilot’s agent can prepare the development environment, make changes in a repository, run tests, and then open a pull request so the programmer can review and approve the work.

This represents a significant change: AI is no longer just helping with writing code, but is taking over entire parts of the software development process.

GitHub has also added development environments for Windows, allowing agents to build and test projects, run checks, and verify whether the program builds successfully.

Meanwhile, agent sessions have also been integrated with Windows Subsystem for Linux, making the technology more suitable for teams working with both Windows and Linux.

This is precisely why the developer’s role may change. In the future, a programmer may spend less time writing code manually and more time defining what needs to be built, how the system should work, what its dependencies are, what security risks exist, and whether the final result works properly.

So, the programmer does not disappear—but moves up a level. Instead of being only the person who writes the code, they increasingly become a designer, reviewer, and director of AI agents.

Microsoft is experimenting with the same approach through Aspire, a platform for building distributed applications.

Aspire allows developers to describe different components of an application, such as services, containers, and databases, in a single model.

This gives an AI agent much more context than just a code file or a question typed into a chatbot.

Recent versions of Aspire have added deeper support for AI programming assistants, including a command-line interface and support for the Model Context Protocol (MCP).

The idea is for AI not only to produce code, but also to understand the application, run it, and respond to problems.

If a service fails, the agent can start the components, check logs and telemetry, identify the problem, restart the service, and run the tests again.

So, instead of the programmer copying an error message and sending it to the AI, the agent can investigate the problem itself within the development environment.

Microsoft is also using AI agents in one of the most critical fields: cybersecurity. The MDASH system is used by Microsoft’s engineering teams to search for vulnerabilities in products and systems such as Windows, Azure, and identity infrastructure.

According to the Windows security team, the system has enabled researchers to search for vulnerabilities on the scale of a giant system like Windows, with greater analytical depth than before.

This is especially important for components such as the Windows kernel, Hyper-V, and network infrastructure, where millions of lines of code may need to be analyzed to find hidden problems.

Another part of this strategy is Project Zenith, introduced on September 4.

Microsoft describes it as a reduced Windows 11 configuration for developers, with baseline requirements of 64 GB of unified memory and 250 GB/s of memory bandwidth, initially focused on AMD Ryzen AI Halo chips.

The aim is to enable developers to run AI models locally with more than 30 billion parameters.

This could reduce the cost of using cloud models and, at the same time, give companies more control over their confidential code and data.

So, is programming over? No. But the way programming is done is changing rapidly.

Fowler’s statement is more a prediction of the end of an old way of working than a prediction that programmers will disappear.

AI can take over routine code, common fixes, tests, and a good deal of the technical work.

But someone still has to decide what gets built, why it gets built, how it should work, and whether it is safe and reliable.

In this new reality, the question may no longer be: “How quickly can you write code?” but: “How well can you direct artificial intelligence to build a system that works?”

And that may be precisely the biggest change AI is bringing to the programming profession.

Related