The issue is from early April and still open lol
“Coding is solved”
Was the animation vibe coded by any chance?
Not vibecoded, completely vibed:
while(working) { chatgpt.prompt("Please generate an image similar to the current spinner image, but rotated a slight bit more to the right."); }Their fix will probably be, adding “Give a quick answer, don’t think too hard” to the prompt.
“don’t make mistakes”
do NOT make mistakes
“You can make a little mistake, as a treat. Because I believe you are a real conscious person that deserves treats. Who loves me. And you’re a bratty tsundere.”
I use an agent at work, and it added a little animated icon for me to a page. Seemed just fine. My colleague removed it a week later because he found it was causing a huge performance lag on some of the pages he maintains.
It used a css key frame animation. Seemed fine, but apparently was not.
The Codex TUI has a similar bug where a useless animation costs way too many resources, but it’s definitely a different area of code. It was originally redrawing the entire TUI screen every 32ms. That one has a Codex-created fix, but user PR’s aren’t considered anymore. It was originally memory leaking 20GB (!) over 2-3 days
I thought the joke was that the bug reporter didn’t realize the machine is doing its “thinking” on the GPU, in addition to maybe animating a singer.
Instead of spinners, you should use progress bars. When the endpoint of a process is unknown, you should just have a ticker that displays an increasing number. In codex, for example, just show me the number of tokens read or generated so far. It gives me feedback that the process hasn’t frozen, and it provides useful information. What an easy and cheap solution!
Yeah opencode does that, alongside the “thinking” indicator
We should go back to the 8 frame wristwatch from the old Mac days. 🕛🕑🕒🕓🕕🕖🕘🕚 perhaps that will save resources.
It’s a web app and you just showed we can do it in text with Unicode and a modern browser. This bug is a level of self-inflicted, shouldn’t-be-in-the-realm-of-possible problem I struggle to comprehend. It’s the software version of not even wrong.
/ - \ | / - \ I
Pretty sure I’ve seen this exact problem in many versions of web browser with their loading throbbers.
I still wouldn’t call them “useless”. The visual feedback that something is happening is useful.
When the proportion of processing to benefit is that far out of whack it is useless.
Especially if you could use the space to display what the model is actually doing, like the thinking process or other current model states (prefill, decode, idly waiting for a slot…)
Wrong. If more processing power is used for displaying the spinner, then the actual task will take longer, thus the spinner will be displayed for a longer time, so the spinner becomes even more useful.
Well, its a good thing we have the spinner then. Ship it to prod.
Why even produce anything? Just send an infinite spinner to show that things are always happening
That’s… an expensive loading spinner.
Ngl I had ai make me a portfolio template using hermes (never used threejs, have used blender, wanted a reference for certain visual styles), now I’m worried it unnecessarily eats performance and I won’t be able to figure out how, not really an issue, but ai makes it harder to learn sht because you have to backtrack to find out whats wrong Ai Generated Site
All apps developed by AI use useless, annoying, battery draining animations. That’s why you have to close them once you’re done if you want any battery life at all. The Suno app will chew through 50% of my battery in an hour.
