Pragmatic Engineer's 2026 survey of 906 working engineers asked a simple question: which AI coding tool do you actually love using? Claude Code came out on top, rated most-loved by 46% of respondents, ahead of every other tool in the survey.
46% of 906 engineers rated Claude Code the most-loved AI coding tool. (Pragmatic Engineer, 2026)
A most-loved rating tells you engineers enjoy using the tool, which is a real signal, but it's a different signal from "everyone using it is getting the most out of it". The same 2025-2026 METR research covered elsewhere on this site found a large, measurable gap between developers who believed AI tools were making them faster and those who actually were, and there's no reason to think a well-loved tool closes that gap on its own.
The gap tends to come down to context management: whether the tool has enough of the right information about your codebase, your conventions, and the specific task, to produce something worth shipping without heavy rework. That's a skill, not a product feature, and it's the difference between a tool that's fun to try and one that's actually making a team faster.
In practice, the developers getting the most out of Claude Code tend to do a few things consistently: they point it at the real repo instead of a paraphrased description of the problem, they ask for a plan before they ask for a diff on anything non-trivial, and they review its output the way they'd review a capable but unsupervised junior engineer's pull request, not the way they'd rubber-stamp their own work.
None of that is complicated, but it's also not obvious from using the tool alone, which is exactly why watching someone do it well, live, closes the gap faster than reading about it does.
Source: Pragmatic Engineer, "AI tooling in 2026" survey of 906 engineers.