Fastly‘s new survey on developers using AI for coding shows that 90% of developers are now using AI coding assistants like GitHub Copilot, Gemini or Claude. But if you dig into the numbers, the perceived productivity gains may be illusory.Senior Devs Are All In, Juniors Play It SafeThe survey of nearly 800 developers showed that senior developers with over a decade of experience are shipping 2.5 times more AI-generated code than junior developers. The numbers indicate that one-third of senior devs say more than half their production code comes from AI, compared to just 13% of juniors.“Senior developers probably have a better understanding of how to build requirements, how to write prompts, how to work with those tools,” said Austin Spires, senior director of developer engagement, who handles developer outreach at Fastly and has been coding since the early 2000s. “Junior developers, though, I think it’s actually a good thing that they’re writing a little bit less AI code.”Spires calls it “The Kids Are All Right” approach, where juniors need to learn the fundamentals before leaning on AI crutches.“At least when I was learning to code, I was taught to avoid those cheater tools, start with black and white text, learn the code, learn the hard way, and then start building those augmentation layers on top,” Spires told The New Stack.One senior developer in the survey put it bluntly: “AI will bench test code and find errors much faster than a human, repairing them seamlessly.” But a junior developer highlighted the flip side: “It’s always hard when AI assumes what I’m doing and that’s not the case, so I have to go back and redo it myself.”The Great Time-Saving IllusionThe survey showed that 28% of developers say they spend so much time fixing AI-generated code that it wipes out most of the time savings. Only 14% said they rarely need to make changes. Yet somehow, over half still swear AI makes them faster.Spires said AI feels incredibly smooth when you’re using it — code appears with a few keystrokes, giving you that rush of momentum. But then comes the debugging, the testing, the inevitable “wait, this isn’t quite right” moments that eat into any gains, he said.Recent research from outside Fastly backs this up. A controlled study found that experienced developers actually took 19% longer to complete tasks when using AI tools.“When you think about senior developers, most of their job is talking to people, managing requirements, managing stakeholders,” Spires said. “They seem to be super excited about AI tools because they can kick off experiments. They seem to be approaching it more in a scientific hypothesis way, where they’ll have AI run multiple experiments, come back.”He points to the human element here. “If I’m correcting a robot with my side projects, I don’t feel like I have to worry about hurting the robot’s feelings or anything like that. I would just be like, ‘No, that’s not what I wanted,'” Spires said.According to the report, one respondent put it this way: “An AI coding tool like GitHub Copilot greatly helps my workflow by suggesting code snippets and even entire functions. However, it once generated a complex algorithm that seemed correct but contained a subtle bug, leading to several hours of debugging.”Meanwhile, The New Stack Research Director Lawrence Hecht noted that the Fastly report did not identify what percentage of the survey was junior versus senior developers.Hecht also said: “People have to stop measuring the degree to which ‘AI-generated code’ is created/deployed because there is a big difference between using AI-assisted code suggestions and something that has been vibe coded.”More Developer JoyYet, despite questionable productivity gains, nearly 80% of developers surveyed said AI tools make coding more enjoyable. They’re getting unstuck faster, skipping grunt work and apparently having more fun while doing it.“You can’t outsource taste to an AI,” Spires insists. “You need to have good professional sensibilities to know what good looks like from the AI.”That might explain why senior developers are more bullish on AI’s speed benefits. While half of junior developers say AI makes them “moderately faster,” 26% of senior developers claim to be “a lot faster” — double the rate of their younger colleagues.The Environmental ElephantThe survey also dug into something most people are not thinking about: AI’s carbon footprint. Apparently, developers are pretty aware of it, as about two-thirds across all experience levels know these tools consume serious energy.Senior developers also care about writing energy-efficient code. Nearly 80% of mid- and senior-level engineers actively consider energy use while coding, compared to 56% of juniors.Not Replacing Anyone (Yet)Contrary to some tech leaders’ recent claims about replacing junior to mid-level developers with AI, this survey suggests something different. Spires said he did not see any panic in the survey responses.“I didn’t see any response data from junior developers being afraid of losing their jobs, or senior developers being like, ‘I can outsource this completely,'” he said.Instead, Spires said he’s seeing AI as a tool that could democratize coding.“Making technology accessible to more people is, generally speaking, always a good thing,” he said. “If there are people who use something like AI to build their first application, and that’s how they get exposed to bringing an idea to life, that’s how we’re going to find the next batch of infrastructure engineers.”The Real StoryThe Fastly study indicates that AI coding tools are not the productivity miracle they were pitched as, but they make developers happier. Senior developers are using them more aggressively because they have the experience to catch AI’s mistakes. Junior developers are being smart about learning fundamentals first.Fastly surveyed 791 professional developers in July 2025.The post Fastly: Senior Devs Ship 2.5x More AI Code Than Juniors appeared first on The New Stack.