There does seem to be a bug - if you drag the tongue to the bottom left-most position, close the lips and bring them back as far as possible, raise the larynx all the way, and lower the soft palate all the way, the synthesizer starts making some kind of inhuman screeching sound (which is definitely not physically accurate).
My best guess for why is that the simulation parameters are set up slightly wrong in a way that is mostly fine for normal simulations (not an audible amount of distortion) but quickly goes off the rails on weird cases like the aforementioned.
Overall a really nice app though; honestly I was kinda blown away by it.
Apple is paying API prices, not subscription prices, though. $6000/month is half the engineer's monthly salary, it seems ridiculous to say that won't be enough, unless you're being very inefficient with token spend.
Yeah. I think people are reading $200 as "per month" and not "per day". I do a lot with unrationed Claude tokens and $6k/month is fine. Have I exceeded it? Yes. But it's very rare.
Things that drive up spend is mostly doing other people's work. Triage, "that doesn't seem quite right, let's see what happens if I prompt it", properly investigating incidents (I get a lot where the reporter's agent didn't read the code and just guesses at how it works, kind of making the report useless). Fixing those things is what would really get token spend down -- a consistent way of making bug reports against other teams that has the same context as the people that work on it everyday. I don't know if it's possible. But I'm anti-rationing because the first thing you do when your tokens are counted is to stop helping other people, right?
The fact that it's expensive has nothing to do with how useful it is though... I know plenty of people who max out their $200 plans every week, and they are working on stuff a lot less financially valuable than what Apple employees work on.
You caught me. I would guess though, that at the end user/retail level, more money is spent on lottery tickets than AI subscriptions. Does that work better?
A MacBook Pro is one of the most performant possible computers you could try this on. Electron apps are absolutely terrible without those kinds of resources (RAM especially). Keep in mind that Discord is also an app intended to be used by people who are gaming or streaming video as it’s original purpose, so it should be using as little CPU as possible to not interfere with e.g. the game.
More than that: it's a full browser that doesn't share any of the system browser libraries (using it's own browser engine) and also NodeJS. You can often save almost ~300-500 megabytes by switching to something like Tauri (even Tauri+NodeJS) that uses the system's own browser framework.
Well, Discord is a pretty complex app. It’s loading/caching images, subscribing to many chats at once, needs to be able to hop into a call at any time, and also renders (most) arbitrary Markdown.
Well, keep in mind that you’re not expected to have the entire UI in one statement. You’re supposed to break this up into functions. That’s why stuff like Flutter and SwiftUI work.
Also, this is basically what Tailwind code looks like.
Just tossed the clanker at some stuff in Cyberpunk 2077, seems to be reversing it fine. Mind you I’m using my Claude sub in OMP so maybe I’m getting a less shitty system prompt (also I have the defensive CVP which presumably allows a lil more rev work than normal, though it’s still not enough for interesting things on Opus 5.5)
It’d probably be trickier with multiplayer games given the slopmachine would probably assume you wanna cheat in it
reply