InsightsEngineering4 min read
Choosing a framework in 2026: Next, SvelteKit, or Remix?
All three are good. Defaulting to the wrong one will still cost you a quarter of work you didn't budget for.
Published 17 May 2026
There is no universally correct answer here. There is a correct answer for each of three buyer shapes, and the cost of picking the wrong one is real. We have shipped production work on all three in the last twelve months; this is how we route a new project among them.
Next.js when the surface is dense and the team is many
Next 16 is the right default for content-heavy, SEO-led products with multiple teams contributing across the same codebase. The App Router's server-component model maps cleanly to the way real product orgs split work (each team owns a route group). Vercel deployment is a non-event. Image and font optimization are baked. The ecosystem is the largest by a wide margin.
The cost is complexity. Next's mental model has more moving parts than the other two: server vs. client components, server actions, middleware, route handlers, edge vs. node runtimes. A small team writing a small product spends a lot of bytes learning Next before shipping anything.
SvelteKit when the team is small and the product is opinionated
SvelteKit is the right default when one or two developers own the codebase, the product has a strong opinion about its own UX, and you want the codebase to stay readable as it grows. The component model is genuinely simpler, less ceremony per component, less library churn, smaller bundles. Reactive-by-default makes state management trivial for medium-complexity apps.
The cost is ecosystem. Some libraries we depend on regularly have no first-class Svelte version, only a port that lags a release behind. Hiring is harder; the SvelteKit talent pool is one-tenth the Next.js pool.
Remix when the form is the product
Remix (now merged into React Router 7) is the right default for products where the dominant interaction is forms: admin tools, internal apps, CRUD-heavy SaaS dashboards. The framework's bias toward HTML form submission, progressive enhancement, and nested-route data loading matches that shape better than Next's bias toward server components. We have one client's entire ops console on Remix and it remains the cleanest codebase in the portfolio.
The cost is the cultural moment. The Remix-to-React-Router migration left docs and community knowledge split across two eras, and the deployment story shifted twice during the transition. If you need stability above all else, Next.
Related insights
All insightsEngineering4 min
Why we standardize on TypeScript across every project
Not because of the types. Because of the refactor-without-fear and the speed at which a new developer becomes useful.
Engineering5 min
WebSockets vs Server-Sent Events: when each one wins
Two protocols, two cost models, two operational profiles. The default answer is not always WebSockets.
If this post resonated, here is the work it connects to.
Want to run the numbers for your team?
30 minutes to do the math on your actual roadmap, including when the answer is not Stacklane. No pitch deck.