Microsoft asserts coding remains a vital skill despite AI's growing role in software development

Microsoft Learn, the company's official training and certification arm, dropped a single line on X this week that walks straight into the most anxious question in tech education right now. "If you're wondering whether it's still worth learning how to code, the answer is: Now more than ever." No hedge, no asterisk, no follow-up thread explaining what it really meant. Coming from Microsoft, the timing reads as either reassuring or slightly cheeky, depending on where you sit. This is the same company whose CEO told a room at Meta's LlamaCon that 20% to 30% of the code sitting in Microsoft's own repositories was already "written by software". It is also the company whose CTO, Kevin Scott, has predicted 95% of all code will be AI generated by 2030. So the message to students is not that nothing has changed. It is that the change makes the skill more valuable, not less.

The numbers behind the anxiety are real and they keep moving. Nadella noted the AI share was climbing steadily, and that results varied by language, with more progress in Python and less in C++. Google has gone further. Sundar Pichai said in April this year that 75% of all new code at the company is now AI-generated and approved by engineers, up from 50% the previous autumn. Anthropic puts its own figure above 90%. That is the backdrop for every computer science student watching placement season and wondering if they picked the wrong branch. Entry-level hiring has tightened. The "write this simple function" interview question has aged badly. And a lot of loud voices online have taken those two facts and concluded that the whole path is closing.

Microsoft's argument, and the one most working engineers make, is narrower than the headline suggests. AI is good at producing code. It is not good at knowing what to build, spotting where something will break in production, or taking responsibility when it does. Somebody has to read the output and decide whether it is right. You cannot do that without knowing how code works. There is also volume to consider. More code is moving through every project than ever before, which means more review, more debugging, more architectural calls. The bottleneck moved. It did not disappear. What has genuinely changed is the shape of the skill. Memorising syntax was never the interesting part, and it is now close to worthless as a differentiator. Understanding systems, reading unfamiliar code, and directing AI tools well are where the leverage sits. Stack Overflow's developer survey found 62% of developers were already using AI coding tools daily, up from 44% the year before. Adoption is not a trend anymore. It is the job. Microsoft Learn, for its part, offers free structured paths in Python, C#, and cloud development, which makes the post part encouragement and part pitch. Both things can be true. For a student staring at a compiler error at 2am, the answer stays the same: keep going.