INTELLECT-2, "The First 32B Parameter Model Trained Through Globally Distributed Reinforcement Learning", also thinks Svelte 5 doesn't even exist yet and only speculates about its features

(this is from their website, not run locally like the previous examples)

gemma-3-27b-it also basically makes things up. Just imagine what these things tell you about concepts you DON'T understand...

deepseek-r1-distill-llama-8b just hallucinates complete nonsense

Here's an example of what I would currently prefer to do imperatively instead of declaratively using $effect like it's done in their own example in the documentation. I would call a function that updates the scroll position wherever the messages are updated instead.
svelte.dev/docs/svelte/...

I'm finding it impossible to explain this in under 300 characters, I'd have to write an long-form article.

Also, I'm not saying anything revolutionary here, I think, it's just that I got fooled into too much reactive, functional code without understanding the value of imperative code first.

I'm building 3D animations, I'm using Threlte for some of it, and it has STILL proven valuable to simply load the GLTFs into the scene and animate them imperatively instead of actually using Threlte's functionality as intended. I might eventually collect my thoughts about this and write a bit more.

I don't think this is emphasized to the extent it should be and LLMs consistently generate the spaghetti code and say it's idiomatic to Svelte so clearly other people are not following the Svelte's team advice either

"Helping end world hunger" and "ending world hunger" are two completely different things. Conflating the two things is such lazy thinking...

This discussion started with a guy saying Mark could single-handedly end world hunger, is what you're saying something that Mark could do on his own? No, so this isn't an answer to his question.