I love how many people are attaching to the "half-bounch" startup time metric. I used to do paid macOS dev ~20 years ago and most macOS devs I knew then used the bounce metric! Back then multiple bounces was the norm due to HDDs. One bounce really wasn't possible until SSDs. We've lost this obsession and its time to bring it back.

Aug 29, 2026 · 1:59 PM UTC

60
52
2,655
788,996
Sort replies: Relevant Recent Liked
Replying to @mitchellh @JorgeWritesCode
My first thought was “this can’t be a notarised build, right?” just that there’s a lot of security checks that slow down app launches, no matter how lightweight your app is.
2
71
21,349
It is notarized and stapled so it avoids some of this slowdown but yeah.
65
11,216
Replying to @mitchellh
do you think it will be possible in the future to have Superlogical running on linux with the same/better response time?
1
28
19,434
Yes but the UI can’t possibly look as good especially animations because the popular UI toolkits just aren’t there. But in terms of performance and responsiveness yes Linux can actually do better
5
2
108
17,205
Replying to @mitchellh
What does it take to have no bounce. Is it even possible
1
14
15,112
No, dyld, AppKit startup, and initial window loading (fully in Apples court notuing we can do) is like 150ms on high end hardware. A full bounce is around 500ms.
2
108
13,635
Replying to @mitchellh
Measure until you can type, not until the dock stops. An empty window kills the bounce and still leaves a spinner. Half bounce on paper is easy to fake.
1
2
11,002
Yes it’s interactive at half a bounce. I say that in the video.
1
49
10,288
Replying to @mitchellh
Are you counting bounce-stop or first keystroke?
1
1
6,068
First interactivity accepted
1
13
5,475
more replies
Replying to @mitchellh
@grok what does a half-bounce startup time mean?
1
2
1,793
Replying to @mitchellh
I was like “oh he’s OG!” When you said that. Made me jump in my seat lol
4
1,985
Replying to @mitchellh
It is awesome! I love ghostty and look forward to getting my hands on Superlogic.
1,510
Replying to @mitchellh
More people need to think about optimization, reducing user time, binary size, memory footprint. The problem is a lot of devs out there make their decisions based on what is best for them not the users. The needs of the few outweighing the needs of the many is not it.
8
3,614
Replying to @mitchellh
Half-bounch hh? Not Filesmash 🤣😅#filesmash
1
3
688
Replying to @mitchellh
What's a half-bounce?
7
782
Replying to @mitchellh
old mac devs are about to stand around the dock counting bounces like track coaches
1
2,144
Replying to @mitchellh
half-bounce is the only honest macOS bench. seconds lie, the dock doesnt
1
2,101
Replying to @mitchellh
speaking of performance - are you planning on solving typing latency on remote envs? Like mosh-ish?
1
1,138
Replying to @mitchellh
What is a bounce? What does that mean? Is that a macOS thing?
1
649
Replying to @mitchellh
What sort of things do you lack for it to look as good as it does on mac, on linux?
1,189
Replying to @mitchellh
bro please tell me half bounce is the metric now
1
1,495
Replying to @mitchellh
What about Records for Mac?
1
137
Replying to @mitchellh
hw speed increase made devs lazy about optimization
1,435
Replying to @mitchellh
Looks like I need to add boot time optimization to my upcoming feature list... At least I'm faster than After Effects. (I have quite a few heavy dependencies though, like CoreAudio, CoreMIDI, CloudKit - but most of my startup time is in the json processing of my user saved banks
370
Replying to @mitchellh
Seems like something @cmuratori would have something to say about :)
855
Replying to @mitchellh
Looks like speed’s back on the menu, boys!

ALT Lets Go Boys GIF by Pixel Bandits
1,571
Replying to @mitchellh
When is the release? Demo looked fire
1,478
Replying to @mitchellh
it's taken me 20 years to hear about this but i love it already 🤗🤍❤️🤍🥰
523
Replying to @mitchellh
that's a nice exercise in nostalgia, but doesn't change the fact that this metric still doesn't account for the vastly different scale of development today: we're talking about building complex ai models and systems, not just tweaking some cocoa code.
284
Replying to @mitchellh
half bounce can hide deferred initialization; first interaction latency is the better production metric
1
1
187
Replying to @mitchellh
If you open Claude Code you'll realize that devs really got lazy and used to pushing features vs actually building something seamless.
406
Replying to @mitchellh
Software got too complicated (or too bureaucratic) so people decided to not optimize With AI, these possibilities are real again
3
486
Replying to @mitchellh
Bounce count worked because it was the one benchmark users ran without being asked. Still is. What bothers me is where the SSD gains went. We got launch down to one bounce, then handed the whole budget to Electron and web views, and ended up back at multiple bounces on hardware that would have looked like a supercomputer in 2006. Same metric, same result, three hardware generations later. I don't think users got anything out of that trade.
3
746
Replying to @mitchellh
The bounce catches what profilers miss the time before your code runs. Flutter macOS has a VM init phase invisible to DevTools that eats a full bounce regardless of how fast your first frame renders.
1
1
526
Load more