Last year I published here a blog post about My React’s ecosystem stack in 2025, One year has passed and the good news is that little things have changed. I strongly recommend that you read that post before continuing, and let’s dive in 2026.
First, the ones that did not change:
- Zustand: it grew even more this year, becoming the default state manager.
- Tanstack Query (previous React Query): continues to be a strong option to server state, the experience of using it is smooth and after some time, you start to think how you could not use it for so long.
- Tailwind: simply dominates the frontend area, with a very good usage with LLMs, the experience overall is good (in the beginning I didn’t like it, but then I started to like after some time using)
- Shadcn/ui: the growth of LLM usage brought the necessity of easy to use and customized UI component library, shadcn/ui was in the right place at the time, it’s the de facto default of the industry today. The death of styled-components backlashed css-in-js, so tailwind and shadcn/ui just had an easier time dominating the industry.
- Vitest: for new projects, to use jest feels obsolete. In the last 2 years I have been using vitest and the performance and usability is awesome.
- React Testing Library: still the incomparable best tool for React component tests.
- Vite: in the bundlers world, vite still dominates the frontend industry.
- React-hook-form: still the best tool for form state management, using it with Zod, Yup or other data validation tool makes it easier to use.
But some things changed this year as well, some for the best and some for the worse.
React Server Components
2025 was not an easy year for RSC, the current perception is still not as good as they expected, even with Dan Abramov’s articles about it, that helped clarify their ideas for more and more people.
In my perception, the struggle RSC has it’s more about their intentions and communication. In JSX Over The Wire Dan writes about the ideal usage of RSC. And, contrary to what people used to think, RSC doesn’t challenge SSR, it challenges Rest APIs, and it makes a huge difference.
Until now, the React team used to talk with front-end people, which strongly resonated over the years. But now, if they want to change the way people structure their apps in replacement of Rest API, they need to talk with a different audience (tech leaders, architects, management), because this change doesn’t just affect the UI side, but the whole system architecture.
The problem is: They haven't been able to really communicate with this new audience.
So, the perception of people is really strange, some of them confounds it with SSR, others critiques the “complexity”, others don’t see a strong value to stop using SPA to embrace it, and some of them have real trust issues with Vercel (the company backing Next.js and the React’s team RSC model implementation).
And, we had what happened in December. Several critical CVEs happened in RSC implementation, needing a fast response from the React core team, you can see more in this blog posts:
- Critical Security Vulnerability in React Server Components
- Denial of Service and Source Code Exposure in React Server Components.
There are some things it’s important to clarify:
- CVEs are obviously bad, but they happen (not just with React, by the way).
- The response of the React team was very good, responsible, fast and honest.
- People used these issues to say a lot of misinformation about the RSC.
- In the end, this case of course hurts RSC reputation, even being a normal thing to happen.
Next Framework
In terms of usage, Next is still the bigger React based framework, but we are seeing more and more discontent with their paths, their management (especially with polemics involving Vercel) and their bet in RSC (connecting with last topic).
I’m seeing some people changing to Tanstack Start as well, so I put it as the possible next big tool in the React ecosystem, due the good history of Tanstack and their ability of communicating and creating good tools with the learnings of the past.
And talking about Tanstack…
Router
Over the last seven years, I have used React Router, until they reached version 7 and did that mess with library/framework mode, Remix, etc. They seem very confused, changing directions all the time, creating a bloated library that has one of the most confusing documentations in the ecosystem.
Then, I tried the Tanstack Router, a new router created by Tanstack with the concepts of cache (coming from React Query). And, if you read the article until here, it is obvious that the experience is good. It integrates very well and it allows the usage of the Route Loaders concept.
As I normally work with SPAs, Tanstack Router brings some important tools to evolve the usage, performance and usability of it. I really recommend people to try it this year. The Future For the ecosystem, I see nice component libraries alternatives to shadcn/ui as Base UI, HeroUI and an animation library called React Bits growing strongly.
Animation will be a great topic this year, with React releasing View Transition in some new minor version of v19. During the year, more and more people will update to v19, enabling new features, especially Activity and concurrent features (I pretend to write some articles about it).