What I want to learn in 2025

Change is the one constant, but I think frontend development changed more in the last year than it did in the 5 previous. Here's what I'm looking to get into this year:

1. WebAssembly (Wasm)

WebAssembly has matured significantly and is becoming increasingly mainstream. I want to explore how to leverage Wasm for high-performance web applications, particularly focusing on running complex computations directly in the browser. I'm especially interested in building applications that can run both in the browser and natively, while maintaining security through local-first processing of sensitive data.

2. Svelte

Svelte's unique approach to reactivity and its minimal runtime footprint make it an intriguing framework to learn. 2024 was the year where I was finally able to get out React world and explore more. This year I want to go deeper on svelte.

3. Tanstack Start

I've been following Tanner Linsley for a while now. I went deep into react query in 2023, and have been watching the progress Tanstack has been making with their router implementation and now their meta-framework Tanstack Start. I'm particularly excited about their opt-in approach to server-side rendering.

4. Web Components with Shadow DOM

Web Components are gaining more traction as a native way to build reusable UI components. I'm particularly interested in learning how to build framework-agnostic components. I want to learn how they can work with modern frameworks.

5. Vinxi

Related to #3, I want to explore building my own meta-framework powered by vinxi. This is what tanstack is using to power Start. There are a lot of great primitives for building both client and server endpoints. I am particularly interested in a framework that could provide a server entry that will return a different bundle based on the user's authorization. This would allow for the dev experience of a single app with multi-app deployment, without having to use something like lerna for shared modules.

6. Bringing AI into the mix

I started dabbling with co-pilot last year, but somehow, the generative auto-completion that seemed revolutionary a year ago now looks quaint. I regularly use Curor's agent to speed up boilerplate but now with things like bolt and replit, I am interested in having AI build simple apps in their entirety. This will allow more bandwidth for design and systems building.