T3 Chat needs an official changelog. I should probably make a page for it. Instead, I'm making a thread for it (It's this thread. All T3 Chat changes will now be logged here)

@Moonytastic @JayzTwoCents omfg can't unsee that now

@thegenioo Yeah we got swamped with a big tech overhaul. DM me your email, you’re getting some free credit 🫡

Thank you to @OpenRouterAI for making it possible to safely deploy DeepSeek R1 :)

Looks like it's a class component that reads it's own content on render and replaces elements in a...kinda sketchy way. Tbh surprised it works at all Reason for issue is that server renders without this step, client gets a different state, they disagree Simplest fix, import it with "dynamic" from next/dynamic, pass {ssr: false} as 2nd argument.

@luciascarlet Do I have an imposter

@fuckgrimlabs I hate the movement of the text in the bottom half but the rest is p cool

@janwilmake @demarc_the_dev 1. Jsdoc has gotten better but still has lots of limitations 2. Browsers will be supporting ts relatively soon. Plugins like ts-blank-space plug the gap fine for now 3. LLMs have way more training on TS by sheer volume. Delusional to think jsdoc will work as well.

@DesignGears They work, tell it to not using single $ syntax (it breaks a ton of other shit) \( and \) work. “Math mode” coming…eventually

@demarc_the_dev @theo 3. an LLM doesn't care if it's JS or TS. just as easy, JS is probably easier to generate.

@r_marked This is your brain on v0

@rex_the_reptile Not viable. The needs of every "local-first" app are so wildly different, a generic would never make sense

@techikansh We're on the latest 4o. Only reason OpenAI says something different is their system prompt

@entrepeneur4lyf For one message, yes

@zeeg Sentry exists because you refuse to use languages with types

@rtaneja_ I am sad we missed this and I'll make @r_marked do it asap