Ask HN: Svelte 5 devs, what's your preferred LLM?

[_jogicodes_]

I was a Claude sonnet boi until pretty much today and that's because Claude tended to beat all all of the other models, usually. However, recently I started working with svelte 5 for which the Claude experience isn't all that great, so today I went back to good old GPT-4o and I was surprised that it did a much better job for me which left me wondering what else I don't know so I decided to ask to swim intelligence of the interwebs:

What's your favorite large language model for Svelte 5 related problems?

[eamag]

I paste this https://svelte.dev/llms-full.txt into https://aistudio.google.com/app/prompts/new_chat and remind to use svelte 5 syntax, it works somewhat ok

[kingkongjaffa → eamag]

> I paste this https://svelte.dev/llms-full.txt

Intrigued, are there there other tools that provide this kind of /llms-full.txt document to load into your system promtpt/use for RAG?

edit: I found this which seems to be similar https://github.com/AnswerDotAI/llms-txt

and

https://llmstxt.site/ to find similar files to give your own LLMs context on various tools

[daemonologist]

I use Supermaven with `llms.txt` in the context; it's passable.

Every model though tends to want to generate Svelte 3/4 syntax. I haven't found a truly good solution, so I'll follow this thread with great interest.

[poetril]

I've been using a mix of Claude and Deepseek via the Kagi assistant to great success. Usually adding relevant sections of the docs alongside my question and reminding it to use svelte 5 syntax.

[trash3]

Have you tried uploading the mini llms.txt for svelte into Claude?

https://svelte.dev/docs/llms