Stack Overflow developer survey 2026: which tools moved since 2023, and which moves are a changed question

Claude Code overtook GitHub Copilot as the most-used coding agent in the 2026 Stack Overflow Developer Survey, published 6 October 2026, with 68.4% of the 11,739 respondents who ticked a tool having used Claude Code versus 61.3% for Copilot — a reversal from 2025, when Copilot led 67.9% to 40.8%. The author's own counts from the public 2023–2026 survey datasets (30,903 valid responses in the newest file) show the coding-agent question changed audience between years, so only the order compares, while editor shares held like-for-like: Visual Studio Code fell from 76.4% to 71.3%, its lowest in four years, and Zed rose from 7.3% to 11.5%.

Stack Overflow developer survey 2026: which tools moved since 2023, and which moves are a changed question In the 2026 Stack Overflow Developer Survey, published on 6 October 2026, Claude Code https://stackness.dev/tools/claude-code passed GitHub Copilot https://stackness.dev/tools/github-copilot on the coding-tool question: 68.4% of the 11,739 people who ticked a tool had used Claude Code, against 61.3% for Copilot. In 2025 the order was the other way round, 40.8% to 67.9%. Two editor moves also hold up when the same option is compared year to year: Visual Studio Code https://stackness.dev/tools/visual-studio-code fell from 76.4% to 71.3%, its lowest share in four years, and Zed https://stackness.dev/tools/zed rose from 7.3% to 11.5%. Most other year-over-year jumps come from questions that changed. The numbers below are my own counts from the public survey datasets https://github.com/StackExchange/Survey/tree/main/packages/archive for 2023 to 2026, 30,903 valid responses in the newest file. Stack Overflow developer survey 2026 tool changes, year over year The 2026 Stack Overflow survey moves that are not question artefacts sit on the editor list, which kept most of its options and wording from 2025. The coding-agent list changed audience, so only the order inside it compares with 2025, not the shares. The table shows the four editor years for the options listed each time. | Editor or IDE | 2023 | 2024 | 2025 | 2026 | 2025 to 2026 | |---|---|---|---|---|---| | Visual Studio Code | 74.3% | 74.2% | 76.4% | 71.3% | -5.1 | | Notepad++ https://stackness.dev/tools/notepad-plus-plus | 24.7% | 24.1% | 27.6% | 28.0% | +0.4 | | Visual Studio https://stackness.dev/tools/visual-studio | 28.7% | 29.6% | 29.2% | 27.8% | -1.4 | | Vim https://stackness.dev/tools/vim | 22.5% | 21.7% | 24.4% | 27.3% | +2.9 | | IntelliJ IDEA https://stackness.dev/tools/intellij-idea | 27.0% | 27.0% | 27.3% | 24.4% | -2.9 | | Cursor | | | 18.0% | 15.0% | -3.0 | | Neovim https://stackness.dev/tools/neovim | 12.0% | 12.6% | 14.1% | 14.8% | +0.7 | | Zed | | | 7.3% | 11.5% | +4.1 | | Windsurf https://stackness.dev/tools/windsurf | | | 4.9% | 3.2% | -1.7 | | Emacs https://stackness.dev/tools/emacs | 4.7% | 4.2% | not listed | 5.3% | | Share of respondents who ticked at least one editor as used. Bases: 85,864 2023 , 57,592 2024 , 26,019 2025 , 13,120 2026 . Stack Overflow Developer Survey data, ODbL 1.0 https://opendatacommons.org/licenses/odbl/1-0/ . Stack Overflow's results site https://survey.stackoverflow.co/2026/technology shows slightly lower figures VS Code 70.9%, Claude Code 65.5% , because it divides by everyone who ticked an option in either the "used" or the "want to use" column. I divide by people who ticked a "used" box, the same base Stackness uses for the survey history on its tool pages. Claude Code passed GitHub Copilot on the coding-tool list Claude Code leads GitHub Copilot in the 2026 Stack Overflow survey by 7.1 points, 68.4% to 61.3%, followed by OpenAI Codex https://stackness.dev/tools/codex-cli at 30.8% and Cursor https://stackness.dev/tools/cursor at 26.6%. A year earlier the same two tools sat 27 points apart the other way. The size of Claude Code's jump, from 40.8% to 68.4%, is not a like-for-like number, because the question changed audience. The 2025 list, "out-of-the-box agents, copilots or assistants", was shown only to people who said they used AI agents daily, weekly or monthly: 8,332 people ticked a tool. The 2026 question, "Have you used any of the following coding agents or assistants in the past year?", had no such gate, per the 2026 question flow https://github.com/StackExchange/Survey/tree/releases/2026/questions . What survives the change is the order. Copilot led Claude Code by a ratio of 1.7 to 1 in 2025, and Claude Code leads Copilot by 1.1 to 1 in 2026, among people answering the same list in the same year. | Coding agent or assistant, 2026 | Share | |---|---| | Claude Code | 68.4% | | GitHub Copilot | 61.3% | | OpenAI Codex | 30.8% | | Cursor | 26.6% | | Google Antigravity | 16.7% | | Gemini Code Assist | 15.2% | | OpenCode | 14.2% | | JetBrains AI https://stackness.dev/tools/jetbrains-ai | 13.4% | Base: 11,739 respondents who ticked at least one tool as used. VS Code fell to 71.3%, and the drop holds among professionals Visual Studio Code's share fell 5.1 points between the 2025 and 2026 Stack Overflow surveys, to 71.3%, after three years between 74% and 77%. The 2026 sample is smaller and drawn differently, so I checked the drop on the closest group the two years share: "I am a developer by profession" in 2025 and "I create code as a primary part of my job" in 2026. In that group VS Code went from 76.5% to 71.9%. The wording changed a little: 2025 asked about "development environments and AI-enabled code editing tools" behind a yes/no screener, 2026 about "development environments and code editors" with no screener. Neither change explains a loss for VS Code in particular, and the options around it moved both ways: Vim gained 2.9 points, IntelliJ IDEA lost 2.9. Among the editors and IDEs developers use https://stackness.dev/categories/editor-ide , it is a real but modest shift in a self-selected sample. Zed gained the most of any editor that kept its place on the list Zed went from 7.3% to 11.5% of editor answers between 2025 and 2026, a gain of 4.1 points and the biggest among options that were on both lists, and the professional-only subset shows the same move, 7.5% to 11.8%. With 13,120 people answering in 2026, the gap is far outside sampling noise, though the sample is self-selected and skews towards people who still visit Stack Overflow. Cursor looks like a faller on the editor list, from 18.0% to 15.0%. It is not: in 2026 Cursor is also an option on the coding-agent list, at 26.6%. Counting anyone who ticked Cursor on either list, it went from 17.4% of people who answered the editor or agent question in 2025 to 22.7% in 2026. Given the choice, some Cursor users filed it under agents. Four jumps in four years come from a changed question Four of the biggest year-over-year moves in the Stack Overflow developer survey since 2023 come from changed wording or answer lists, not from changes in use: GitHub Copilot's fall in 2024, OpenAI Codex's rise, Claude Code leaving the editor list and the switch from models to vendors. 1. GitHub Copilot fell from 85.2% to 42.8% between 2023 and 2024. In 2023 Copilot was on a list of AI developer tools. In 2024 Stack Overflow merged that list with AI search tools, so Copilot shared the question with ChatGPT, which 85.3% ticked. 2. OpenAI Codex went from 1.3% in 2024 to 30.8% in 2026. The 2024 option predates the Codex CLI https://github.com/openai/codex , which OpenAI released in April 2025. The name is the same; the product is not. 3. Claude Code left the editor list. In 2025 it was an editor option at 9.8%. In 2026 it is only on the coding-agent list. Any chart that follows Claude Code's editor share shows it vanishing. 4. The model question became a vendor question. 2025 asked which "LLM models for AI tools" people used "for development work"; 2026 asks which "LLM model vendors" they used "for work". Anthropic at 76.2% and OpenAI at 75.9% in 2026 cannot be set against "Anthropic: Claude Sonnet" at 43.4% in 2025. The headline AI-use figure changed too. 2023 to 2025 asked whether you use AI tools "in your development process"; 2026 asks about AI tools or agents "in your role at work", as a multi-select. Stack Overflow's own retrospective https://stackoverflow.blog/2026/10/01/a-look-back-before-we-look-forward-a-developer-survey-retrospective/ gives 44%, 62% and 79% for 2023 to 2025; the 2026 question puts 17.2% at "I don't use AI tools", and that is a different measure. Tools that dropped off the 2026 lists, and tools that came back The 2026 Stack Overflow survey dropped five options from the editor list: Claude Code, Aider https://stackness.dev/tools/aider , Cline https://stackness.dev/tools/cline and Roo https://stackness.dev/tools/roo-code , Bolt https://stackness.dev/tools/bolt and Lovable https://stackness.dev/tools/lovable . Aider and Cline moved to the coding-agent list, where Claude Code already was, and Roo was dropped; Bolt and Lovable moved to a new list of no-code agent builders. None of the five lost users because of the survey; they changed question. From the 2025 agent list, ChatGPT https://stackness.dev/tools/chatgpt , Google Gemini https://stackness.dev/tools/gemini , Microsoft Copilot and Perplexity https://stackness.dev/tools/perplexity are gone, since the 2026 list covers coding tools only. Emacs, CLion https://stackness.dev/tools/clion , GoLand https://stackness.dev/tools/goland , NetBeans https://stackness.dev/tools/netbeans and RubyMine https://stackness.dev/tools/rubymine came back to the editor list after a year off it. Emacs is at 5.3%, close to its 4.7% and 4.2% of 2023 and 2024, so its 2025 absence was a list cut. New this year: GitHub Codespaces at 6.7% and Kiro https://stackness.dev/tools/kiro at 2.8% of editor answers, and Google Antigravity https://stackness.dev/tools/google-antigravity at 16.7% and OpenCode https://stackness.dev/tools/opencode at 14.2% of coding-agent answers. The survey against npm downloads and the JetBrains survey npm downloads and the JetBrains Developer Ecosystem survey agree with the Stack Overflow survey on direction and disagree on size. In npm's download counts https://api.npmjs.org/downloads/point/2026-10-22:2026-10-28/@openai/codex for 22 to 28 October, @openai/codex went from 288,903 in 2025 to 27,029,722 in 2026, @anthropic-ai/claude-code from 4,906,294 to 12,837,062, and opencode-ai from 50,669 to 4,458,157. Downloads count CI runs and reinstalls, and Claude Code's setup docs https://code.claude.com/docs/en/setup recommend a native installer over npm, so npm undercounts it. The JetBrains survey of more than 15,000 professional developers https://blog.jetbrains.com/research/2026/08/ai-coding-agent-adoption-2026/ , fielded May to July 2026 and reweighted, has the same order at the top, Claude Code 39% at work against Copilot 21%, Codex 16% and Cursor 12%. JetBrains sells competing tools, and its shares are of all respondents, not of people who ticked a tool. Stackness's own trending page https://stackness.dev/trending tracks what members add. On Stackness public stacks, as of 30 October 2026, Claude Code is on 8 real profiles, Cursor on 7, VS Code on 5, OpenCode on 3, OpenAI Codex on 2 and GitHub Copilot on 1, with bot accounts removed how I count https://stackness.dev/about/data-sources . Those are small numbers and show only that the same tools turn up. Method and sample The 2026 Stack Overflow survey has the smallest sample of the four years: 30,903 valid responses, against 89,184 in 2023, 65,437 in 2024 and 49,009 in 2025, per Stack Overflow's methodology pages for 2026 https://survey.stackoverflow.co/2026/methodology , 2025 https://survey.stackoverflow.co/2025/methodology and 2024 https://survey.stackoverflow.co/2024/methodology and its 2023 results https://survey.stackoverflow.co/2023/ . It ran from 23 June to 5 August 2026 and was promoted on stackoverflow.com, the blog and an opt-in email list, so it over-represents people who still use Stack Overflow. - Bases. Each share divides the people who ticked an option as used by the people who ticked any option as used on that question. Empty answers are left out; "Other" stays in the base. - Who answered. Only 13,120 of the 30,903 answered the 2026 editor question and 11,739 the coding-agent question; 65.5% of the 14,347 who answered the closing question on length called the survey too long, per the archive's methodology data. - Bots. The 2026 validity rule keeps some responses flagged as bots if they answered ten core questions, one of them the editor question. - Stackness history. The survey series on Stackness tool pages run to 2025. The 2026 dataset was published on GitHub on 13 October under a new file name, and the counts in this post come from that file. Key numbers - 68.4% of 2026 coding-agent answerers used Claude Code, against 61.3% for GitHub Copilot Stack Overflow Developer Survey 2026, published 6 October 2026 , base 11,739 . - 40.8% Claude Code and 67.9% Copilot on the 2025 agent list, shown only to AI agent users 2025 survey, base 8,332 . - 71.3% of 2026 editor answerers used VS Code, down from 76.4% in 2025 base 13,120 ; 71.9% among people who code as their main job base 10,620 . - 11.5% used Zed in 2026, up from 7.3% in 2025. - 30,903 valid responses in 2026, against 49,009 in 2025 and 89,184 in 2023. - 27,029,722 npm downloads of @openai/codex from 22 to 28 October 2026 , against 288,903 a year earlier. New posts by email About one email a week with what we published. No account needed. Which of these do you use? Nobody on Stackness runs Notepad++ yet. Tick what you use and be the first. Play Lonny Eagle Amateur Skater 4A pixel art skateboarding game. Signing up opens it straight away.Sign up to play https://stackness.dev/games/lonny-eagle