Microsoft engineer David Fowler said that manually entering code is becoming a thing of the past, and the role of traditional programmers is rapidly changing. In an era of artificial intelligence agents like GitHub Copilot and updated versions of the Microsoft Aspire platform that support the MCP protocol, experts are increasingly involved in setting up tasks, debugging, and validating results rather than writing code from scratch.
However, the new approach revealed a serious shortcoming: the company’s own research showed that code generated by neural networks contains 1.4 times more serious errors and up to 45% more potential vulnerabilities than humans. Although artificial intelligence already writes much of the software for tech giants, the need for technical staff to oversee quality control and architecture remains.