Thread

1. pier25 · 1790950737 · focus · HN ↗

   I used Svelte daily as my main framework between 2020 and 2024. Took me a couple of years to realize Svelte's biggest strength is also its biggest issue. Having a custom language means so much effort is needed to build and maintain custom tooling to support it. And not only from the small Svelte team and its community. Eg: Even to this day, the Svelte plugin by JetBrains is trash [2]. If you want to do Svelte you're basically trapped into using their VSCode extension.

   Ultimately I ended up moving away to Solid primarily because I can write vanilla TS/TSX. Like it or not, TS/TSX is a standard these days and most editors or IDEs support it out of the box.

   [1] <a href="https:&#x2F;&#x2F;plugins.jetbrains.com&#x2F;plugin&#x2F;12375-svelte" rel="nofollow">https:&#x2F;&#x2F;plugins.jetbrains.com&#x2F;plugin&#x2F;12375-svelte

   1. philipthetenth · 1790951661 · focus · HN ↗

      I use Svelte heavily since 2023 and have never even opened VSCode once in my life.

      1. pier25 · 1790953271 · focus · HN ↗

         what do you use?

         1. philipthetenth · 1790996378 · focus · HN ↗

            Neovim
   2. afavour · 1790954888 · focus · HN ↗

      Ironically, given the chatter in this thread, I&#x27;d say this is a problem very solvable by LLMs. The Svelte language is well defined, the JetBrains API is well defined. No reason why a plugin couldn&#x27;t be LLM generated.

      1. pier25 · 1790957983 · focus · HN ↗

         Yes but if you&#x27;re using LLMs then surely something with more data available will produce better results?

         1. afavour · 1790980044 · focus · HN ↗

            Seems like a separate issue? I also wouldn&#x27;t assume that. Svelte is faster and smaller than React, even with all the training data in the world there are still non-negotiables.

Unofficial Hacker News client; not affiliated with Y Combinator.