I moved off of Next.js

Theo rewrote T3Chat from scratch counter: 6

"The main thing we do at our startup is to rewrite the whole app every 6 months, but don't worry, this time it's the best stack possible"

"I fear not the man who had written 10,000 apps, I fear the man who has written 1 app 10,00 times" - Sun Szu

Theo, great video as always. Wanted to add some context on the Start -> Nitro mitigation since I don't think the full picture made it to you before recording. The actual root cause for all of the root issues you were experiencing was a Convex configuration issue: a subscription setup function was running on the server that's designed to be client-only. Convex assumes that function runs on the client, so when it ran server-side, subscriptions were never cleaned up. That's what caused the file descriptor limits and crashes. The fix was a **single line change**: don't run that Convex function on the server. Start itself wasn't the problem, and there's no inherent overhead issue with it for large apps. I get why you'd stick with Nitro after going through all that debugging pain. No shade there. But I wanted to clarify for anyone watching who might think they need to avoid Start or hack around it for scale. They don't. Killed me a little when you said "if I had known we would have had to move off Start for API routes to Nitro I would have blocked it" because... you wouldn't have had to. Julius and I found the real cause weeks ago and I'd asked the team to loop you in. Obviously that didn't happen, or just didn't land like it should have.

anything but the mobile app for t3 chat

waiting for when he moves it back to nextjs

Counting til bro changes to SvelteKit

2030: I rewrote t3chat in Web assembly

Tanner's climb to the top will be studied for generations.

one thing I like about theo is how transparent he is with his tech, really cool to see man