JetBrains' AI Pulse survey (May-July 2026, compared against January 2026) shows more movement in six months than the coding-tool market saw in the two years before it. Claude Code's workplace adoption more than doubled, from 18% of developers in January to 39% by mid-year -- 47% in the US specifically, meaning close to half of American developers are now reaching for it, alongside the highest satisfaction score of any tool measured.
OpenAI's Codex posted the steepest growth curve of anything in the survey: workplace adoption went from 3% to 16% in the same window, roughly a five-fold jump, and the standalone Codex product crossed 5 million weekly active users in June. Cursor, meanwhile, slipped -- from 18% to 12% workplace adoption -- even though it remains the most broadly adopted tool among individual developers and small teams outside the enterprise survey population.
Two more data points round out the picture. GitHub Copilot still leads on raw deployment count -- it's bundled into more organizations than any single competitor has sold directly -- while Claude Code and Cursor score highest on the satisfaction metrics that predict whether a tool survives past its first six months. And OpenCode, the open-source entrant with no company marketing budget behind it, has reached 7% adoption and a reported 42% mindshare, which is the kind of number that usually means word-of-mouth inside engineering teams is doing something a marketing budget can't.
The framing JetBrains uses for the year is worth keeping: if 2023 was autocomplete and 2024-2025 was the AI-native IDE, 2026 is the year coding AI moved into agent engineering -- multi-step, tool-using, and judged by developers on whether it can be trusted with a real task, not just a completion.
Satisfaction and adoption aren't moving together this year -- Claude Code and Codex are gaining ground on both, Cursor is gaining on neither -- which suggests developers are actively re-evaluating rather than defaulting to whatever they picked first.