JetBrains just published its 10th annual State of Developer Ecosystem report, drawing on 28,945 developers across 198 countries. The headline: 90% of professional developers now use AI coding agents at work every week, with 68% using them daily. That is not an adoption trend — that is the new baseline. But 46% also say their sense of code ownership has decreased because of AI. The report measures both productivity and something harder to quantify: what it now means to be a developer.
The Numbers Are Not Ambiguous
The 2026 survey data is unambiguous on volume. On average, 47% of all code is now fully generated by AI agents. Over half of developers use agents for more than 60% of their work tasks. One in five writes literally zero code without AI. If your team is still treating AI tools as optional, the data says you are out of step with the profession.
What is less covered is how senior developers are using agents. About 25% of senior developers now generate more than 80% of their code through agents — and they are not just prompting. Experienced engineers give agents structured specifications, verification criteria, and architectural guidelines. Juniors still lean toward AI-assisted workflows rather than fully agentic ones. The counterintuitive finding: experience correlates with more delegation, not less.
Claude Code Dominates; GitHub Copilot Is Losing Ground
The tool landscape has shifted fast. Claude Code is now used by 39% of professional developers — up from 18% six months ago — making it the most widely adopted AI coding tool at work. GitHub Copilot sits at 21%. Codex has grown roughly 5x since January (from 3% to 16%). Cursor is at 12%.
Claude Code is also the most loved tool at 46% sentiment. That matters for teams deciding on standards: when developers choose tools they prefer, adoption friction drops and usage goes up. The enterprise moat that GitHub Copilot built on IDE integration and licensing deals is visibly eroding. If your organization standardized on Copilot two years ago and has not revisited that decision, this is the data to bring to that conversation.
The Ownership Problem
The 2026 report does not hide the uncomfortable findings. Forty-six percent of developers say their sense of code ownership has decreased due to AI use. Twenty-five percent now believe AI will eventually replace developers entirely — up from just 8% in 2024. That is not a fringe concern anymore; it is a quarter of the profession.
The counter-data: 55% say they have become more creative since adopting AI tools. JetBrains' sentiment research shows these two things are simultaneously true. Developers are shipping faster, working on more interesting problems, and feeling less connected to the actual code that ships. Whether that is a good trade depends on how much the craft — not just the output — matters to you.
“The use of AI is now standard practice, not a differentiator.”
JetBrains State of Developer Ecosystem 2026
Read that as both fact and warning. Teams that treat AI fluency as a bonus skill are behind schedule.
Python Finally Overtook JavaScript
After a decade of surveys, Python has overtaken JavaScript as the most widely used language, reaching 65% usage. TypeScript adoption has tripled since 2017 and tops the Language Promise Index — the measure of which languages developers most want to invest in next. Rust is second on that index and leads highest-paid skills alongside Go.
Go posts its highest loyalty score ever: 70% of Go developers say they have no plans to switch languages, up from 64% in 2024. The State of Go 2026 companion report attributes this to the language's fit for infrastructure and AI tooling work — both high-growth areas. The language map has reshuffled; TypeScript and Rust are where developer ambition is pointing.
What to Do With This
If you manage a team: benchmark where you are against 90% weekly use and 68% daily. If your developers are not near those numbers, something is blocking adoption — tooling, trust, workflow, or culture. Find out which.
If you are an individual developer: the senior cohort's shift to giving agents specifications rather than prompts is worth noting. The leverage comes from structured inputs — not better one-liners. Developers generating the most value from AI agents treat them as collaborators with context, not autocomplete on steroids.
The full interactive dataset is available at devecosystem.jetbrains.com. JetBrains also published a deeper breakdown of code generation percentages in August 2026 that is worth reviewing alongside this report.
