Nitor developer survey 2026: AI agents and other tech trends of the year

Now in its sixth year, the Nitor Developer Survey tracks which technologies developers and architects at Nitor have used during the previous year. It asks what they like and dislike, and what they are interested in learning or adopting. This year we grouped the technologies into 18 categories. The survey was open for two weeks in December 2025. It received 96 responses, so the results are representative of the company. In this blog post, I'll share highlights from selected categories and a summary about the results as a whole.

When analysing the results, the three most important metrics are the usage ratio, which is the share of respondents that have used the technology significantly during the past year, the net like ratio, which is the share of users who like the technology minus those who dislike it, and the interest ratio, which is the share of respondents who have not used the technology but are interested in learning or using it. Changes in ratios since the previous years are indicated in percentage points (pp).

AI tools

AI tool usage has grown dramatically across the board. ChatGPT is now used by 83% of respondents, making it the most used technology after company-standard Slack and Microsoft Teams. GitHub Copilot usage has also surged, up 22 pp to 68%. Claude follows at 41% usage, with the highest year-over-year increase (+28 pp) among all technologies in the survey. Gemini usage has also grown substantially, reaching 31%.

Among terminal-based AI tools, Claude Code is the clear leader at 39% usage. This is a striking debut, as it wasn't even included last year. Other terminal-based tools, like GitHub Copilot CLI, Gemini CLI, and Codex, are also new on the list. This reflects how 2025 was the year agentic coding became a realistic option for serious projects – although it still requires solid software engineering skills to get high-quality results.

Overall, AI tools are well-liked. The main exception is Microsoft Copilot, which ranks among the most disliked technologies in the survey (-20% net like ratio). Claude Code was the most liked AI tool, with a 70% net like ratio. Claude itself was close behind at 67%. Cursor was also very highly liked at 75%, although it was listed under editors and not AI tools. The distinction is getting blurrier each year.

The results show that Nitoreans are using a wide array of AI tools to increase their productivity and help with information gathering, coding, and other development tasks. The competition is currently very intense among major AI providers, in particular OpenAI, Google, and Anthropic, so we should expect many changes in the coming year. Mistral is an interesting EU-based alternative, although the survey shows that it has not yet caught on more widely at Nitor. A quarter of respondents have self-hosted models, double the share from last year.

Languages

There are no radical shifts in the languages category, but still some interesting trends. TypeScript usage rose to 80%, making it the most used programming language by a large margin at Nitor. TypeScript only bypassed JavaScript last year, and the gap is already 16 percentage points. SQL remains in third place (57%). Java continues a multi-year downward trend, falling behind both shell scripting (50%) and Python (47%) to sixth place, at 38% usage. Java has nearly halved from 65% in 2021, making it one of the most significant long-term shifts in the survey.

Rust has long been the single most desired technology among Nitoreans, and that is still the case, with a third of respondents interested in learning or using it. Go also shows strong interest and saw its usage rebound to 18%, its best year so far. Kotlin is one of the two most liked technologies in the survey, with a 90% net like ratio. TypeScript and Go are also both highly liked, at 68% and 65% respectively.

JavaScript runtimes

Although Node.js is still dominant in this category (82% usage), with high growth this year, we finally see some movement from one of the competitors. Bun, a JS runtime and toolkit that has long been on the radar, saw usage grow to 10% this year. Bun is one of the two most liked technologies in the survey (together with Kotlin), with a 90% net like ratio. Bun also saw the largest growth in interest of any technology in the survey, moving into second place after Rust.

The other competitor, Deno, also has a very high interest at 19%, which makes it fourth overall. It has not seen any increased usage, though. It looks like Bun may be becoming a real competitor to Node.js. Bun was recently acquired by Anthropic, the company behind Claude, and it is interesting to see what effect this will have.

Frontend

Any rumors about the death of React have been greatly exaggerated. While there is some dissatisfaction, usage still grew to 71%, while the competing frameworks (Angular, Vue.js, Svelte) mostly stayed at last year's levels. Next.js usage also grew slightly (to 40%), although satisfaction has decreased since last year, and is clearly lower than Vite, which is now one of the most liked technologies in the survey. Almost all React metaframeworks (and many non-React frameworks as well) are now based on Vite, with the important exception of Next.js.

Last year, htmx had the second highest interest of any technology in the survey, after Rust. Interest remains high, but usage hasn't budged. Official support for AngularJS (the old 1.x version) ended four years ago, but some poor Nitoreans still have to maintain it in legacy codebases, making it by far the most disliked technology in the survey for the fourth year running.

Testing

Playwright, an end-to-end testing framework, grew significantly to 44% usage. It looks like it might overtake Jest in the top spot next year. Vitest is growing even more quickly, up by 13 pp this year, and is a more direct alternative to Jest. It is also significantly better liked than Jest. JUnit and pytest are both widely used and relatively well-liked in the Java and Python ecosystems.

Cloud platforms

AWS is still the dominant cloud platform. The most interesting change is the significant increase in Google Cloud usage, up by 10 pp to 34%, bypassing Azure for the first time. Google Cloud is also the best liked among the three big cloud providers, with a 52% net like ratio. Another interesting pick is Vercel, which has less usage (14%) but is very highly liked (62% net like). Nitor became a Vercel solution partner in fall 2025, so we anticipate growth here in the future.

Data

The data category, which covers a lot of different technologies, has been relatively stable. PostgreSQL continues as the most popular database, both most used in the category, and very well liked (74% net like this year). Amazon Aurora, a managed relational database compatible with PostgreSQL and MySQL, has seen the largest growth (+8 pp), and is also well-liked (65% net like). S3 and Elasticsearch have also grown notably.

Deployment & infrastructure

In this category, Terraform has been on a steady climb over the years and is now used by 50% of respondents. AWS CDK (30%) and Kubernetes (27%) also saw moderate growth. Both CloudFormation and Serverless Framework are disliked and have been declining for several years. Podman is slowly gaining in popularity as an alternative to Docker, which is still very popular.

Editors

VS Code remains the dominant editor at 79% usage and is well liked (55% net like). IntelliJ IDEA continues a gradual decline, from 74% in 2021 to 51% this year, mirroring the shift from Java to TypeScript. Among AI-native editors, Cursor has grown to 13% usage with high satisfaction (75% net like), and another 13% of respondents are interested in trying it. Zed is also slowly gaining users at 10%. With AI capabilities becoming a core expectation of development tools, this category is likely to see more turbulence in the years ahead.

Communication

In the communication tools category, we see the same phenomenon as previous years: Although both Slack and Teams are widely used, there is a stark difference in how well liked they are. Slack has a very high 77% net like ratio, while Teams is among the most disliked technologies overall in the survey.

Collaboration

Confluence and Jira, both by Atlassian, are widely used but poorly liked. Both see a slight decrease in use, although no clear alternative is emerging yet. GitHub is still very widely used and well liked (74% net like). Trello usage is down significantly due to being internally replaced by Microsoft Planner, which unfortunately almost nobody likes (-47% net like). Figma is extremely popular among designers, and also has a large usage among the developers responding to this survey.

Summary

We calculate positive sentiment as a summary metric using the following formula: (liked + interested - disliked) / total. This gives a good overview of the overall best-regarded technologies. The chart below shows the results across all categories in the survey. The top three items are the same as last year: Slack, macOS and GitHub. These are used by most Nitoreans and very well liked. TypeScript makes a big jump into fourth place this year. The next few items are mostly the same with some noticeable differences. AWS has dropped several spots, and Bun, Claude Code, and Claude are all new on the list, with very big gains in positive sentiment. Compared to other AI providers, Anthropic has focused particularly on developers, and it seems to be making an impact at Nitor. It will be interesting to see if they can keep up the momentum next year.