The problem with this article is that the benchmark method they use is flawed. The documentation of zshbench explains why: https://github.com/romkatv/zsh-bench
Even with a low grade laptop, my zsh config grants me a sub 5ms prompt and a sub 1ms input lag, and that's far more important than the exit time.
./zsh-bench
==> benchmarking login shell of user XYZ ...
creates_tty=0
has_compsys=1
has_syntax_highlighting=0
has_autosuggestions=0
has_git_prompt=1
first_prompt_lag_ms=54.942
first_command_lag_ms=57.069
command_lag_ms=4.275
input_lag_ms=0.669
exit_time_ms=26.522
hyperfine --warmup 3 'zsh -i -c exit'
Benchmark 1: zsh -i -c exit
Time (mean ± σ): 26.5 ms ± 0.5 ms
Range (min … max): 25.5 ms … 27.6 ms
Unrelated to Mise but related to zsh, there's also https://github.com/jeffreytse/zsh-vi-mode/issues/316. I noticed this plugin was causing a lot of delay. Learned a decent amount about zsh profiling from that issue.
I used to use volta but then they killed it and told people to switch to mise. The mise setup is just way too complicated. I just am tired of having different config files for different tooling when it should just read whats in the package.json and be done with it.
This wasn't an option for us because as an org we used their recommended hook (1) to automatically change node versions when switching directories, but it effectively undoes the lazy loading.
With mise you get the behaviour of automatically switching when you change directories effectively for free.
Edit: unless you aliased it to `node` or `npm`, which would be fine I guess but super annoying if you ran node or npm commands often. It is not worth the hassle, no one should use nvm in 2026 imo
How often do you launch a fresh terminal though? I start mine with a script to have favourite tabs ready at boot and then generally not much afterwards.
Constantly. I think we've used the excuse of "well, what if you just launch it less often?" enough to excuse bad performance defaults, especially when alternative solutions fix the issue with very few trade-offs.
I wasn't actually aware of the impact. I measured the zsh startup time locally (with mvn active and commented out) and it indeed makes a difference (.39s -> .08s). Not that I would have noticed that without measuring :) - yes I'm an old geezer.
Thank you for the recommendation, I might then also be able to ditch sdkman as well.
With zsh i set up nvm to lazy load, so I don't pay for it when I don't use it (I'm a C++ and Rust dev, but I occasionally need to run js stuff from other team members).
I can strongly recommend lazy loading in zsh in general, I use it for pyenv too (which is also slow to load, but I write Python maybe every other week or so only).
The way to do this is to use the autoload functionality in zsh and have the autoloaded script replace itself with the real shell init code for the tool in question.
Am I the weird one? I usually have 3/4 terminals open at a time and rarely open new ones. Terminal startup speed is a non-issue for me.
The only thing I demand to be fast on my terminal is grep reverse search (ctrl+r) and of course typing a character. But if your terminal can't keep up with your typing speed there is something deeply wrong with it.
I open and close terminals _constantly_, but I'm usually pretty weird in my workflow, so no comment on your first question.
I run a scrolling WM and have settled into a habit of opening terminals when I need them, then closing them right after. I'll open a terminal, git pull, close it. Etc. I also use a terminal that launches cold in 10-20 ms, so it's not like a pay a price for it.
This is actually what I thought this post was about! But then I saw the Ghostty reference, which, in my experience, is not very fast to launch at all. I got it opening new windows quickly by running the main process as a systemd service, but Foot launches way faster without all that fuss, and allows you to go the daemon route if you want it _even_ faster.
EDIT: Just want to clarify, no shade on Ghostty. That project is cross platform and uses the 100% defensible decision to use the full GTK stack on linux. Foot is Linux AND Wayland only, and uses that very restrictive environment to optimize the hell out of startup and general performance.
I constantly open and close terminals too. Maybe I'm doing a quick lazygit check on cwd. Maybe I'm opening up an ephemeral claude/codex session for a couple questions about why a test failed. Or quickly editing a file with vim. Or remembering where I put that file with yazi or fzf. -- I don't even know, but all of it is contingent on it being fast to open a new terminal in cwd.
So much so that I vibe-coded my own terminal emulator for vertical tabs on macOS (using libghostty for the terminals) that is faster and less weird than iTerm.
It can affect shell subprocess startup time as well, which, depending on your setup and the tools you use, might be worth optimising for.
I don't remember when I did it, but it looks like I must have gone through this at some point (maybe due to using GNU Make a lot? Or perhaps it was some other tool) - my zsh setup does a bunch of autocomplete setup only in the interactive case, and it seems to help a bit with startup time, at least on macOS:
% for i in {1..5}; do /usr/bin/time zsh -i -c exit; done # zsh in interactive mode
0.05 real 0.03 user 0.02 sys
0.02 real 0.01 user 0.01 sys
0.02 real 0.01 user 0.00 sys
0.02 real 0.01 user 0.00 sys
0.02 real 0.01 user 0.00 sys
% for i in {1..5}; do /usr/bin/time zsh -c exit; done # zsh in non-interactive mode
0.01 real 0.00 user 0.01 sys
0.01 real 0.00 user 0.00 sys
0.00 real 0.00 user 0.00 sys
0.00 real 0.00 user 0.00 sys
0.00 real 0.00 user 0.00 sys
For the interactive case, I don't really mind (within limits - the worst case, on macOS after a reboot, takes several seconds, and that's tedious). I also start new interactive terminal sessions fairly rarely.
> But if your terminal can't keep up with your typing speed there is something deeply wrong with it.
For the poor sools that had to work in VDI with radio link...
I open terminal far more often than that. But you should also remember that the startup cost is also paid by some subshells, and any shell scripts you run (the actual cost will vary: which init files are sourced varies between interactive and non-interactive shells as well as login shells and non-login shells, but it won't be zero cost).
Maybe, but notice that almost all of the article, and its followup posted today, is about the speed of starting a shell not of a terminal emulator window. So if you use interactive sub-shells, anything from shelling out of a vi clone or mailx, through :terminal in NeoVIM, to running tmux/screen or script, then the concerns are relevant.
That said, it is almost totally about the Z shell. So you might still qualify as 'weird' in this case on the alternative grounds of not using the Z shell. (-:
I only notice how efficient my terminal is if something dumps a ton of logs to stdout. Sure I can pipe to a file and use vim, but it's convenient not to need to do that. And sometimes I have like 8 things going at once.
Speaking of slow, what I absolutely cannot comprehend is why ghostty is so popular. Despite being written in Zig it is very slow and a total CPU and memory hog. Just sitting there idle it’s pulling a constant 40% of my CPU? No thanks!
I'm the creator of Ghostty. This isn't right. It should idle at 0 to 1%, as supported by sibling comments. If you can collect more details about your system please open a discussion on the main Ghostty repo. Same with memory.
In terms of speed, same thing: if you can provide some kind of objectively measurable thing, we can look into it. Everything we've measured so far firmly places Ghostty in the "fast" camp (with friends such as Kitty).
We're sometimes faster, sometimes slower, but in any case not noticeably so. You wouldn't pick Ghostty vs Kitty for example for performance, it'd be something else. But you would pick Ghostty over say... iTerm2 for performance (but you may pick iTerm2 for features, its extremely feature rich!).
That happened to me too, turns out my desktop was software accelerated because I had screwed up the GPU config somehow. I asked Claude to fix it and it did.
I suspect you enabled some weird setting that you've forgotten about. ghostty isn't unimaginably fast but it's faster than iTerm 2 which is plenty. And I'm sitting here with a lengthy Claude Code session open, as well as a couple tabs for my docker container and dev servers, and its idle CPU usage is 0.0%.
WezTerm is great too and, AFAICT, is everything Ghostty wants to be. Kitty is also very good, although WezTerm edges it out for having an integrated muxer.
I can't understand Ghostty either except as some kind of trendy memetic thing, the $GME of the TTY world.
Yup. For a while I was switching between all three to see which I liked most, and I always end up coming back to Wezterm. The only issue I’ve encountered is that releases on various package managers are routinely broken, but building from source solved that.
I know ghostty is designed to be super high performance but I find it both uglier and slower than Iterm2!
I started using it expecting to love it, but in reality it didn't seem to gain me anything but in fact was worse in several major ways. Also less configurable than Iterm2. :/
$ for i in {1..5}; do /usr/bin/time zsh -i -c exit; done
0.01 real 0.00 user 0.00 sys
0.01 real 0.00 user 0.00 sys
0.00 real 0.00 user 0.00 sys
0.00 real 0.00 user 0.00 sys
0.00 real 0.00 user 0.00 sys
Maybe I should remove the following bit of code from my profile?
if [ "$SESSION_TYPE" != 'remote/ssh' ]; then
if [ "${TERM_PROGRAM}" != 'tmux' ]; then
( if [ $[RANDOM % 10] == "0" ]; then fortune -n 40 -s; else echo "Hi, $(whoami)!"; fi ) | cowsay | lolcat && printf '\n'
else
if [ "${TMUX_PANE}" == '%0' ]; then
fortune | cowsay -f small | lolcat && printf '\n'
fi
fi
fi
It's a whole chain of interpreters firing up (sub-shells, Perl for the cow, Ruby for the lol, I think.) :D
But what would life be without a little bit of fun?
I know nobody is missing it, because it is the first bit of the blog post, but the author does have a follow-up where they note corrections based on push-back they received from a reader.
Apparently for some of the simplicity-produces-speed arguments, users have found complex/featurefull. tools that are still quick. I’m not sure how to evaluate this (I like simplicity just because it is easier to fit simple tools in my head) but we should note the counter argument (and applaud the follow-up).
I read Ghostty runs in a single process, but whenever I tried something like that eg a client/server model in urxvt or foot, I ended up reverting, because eventually some weird state affected the daemon and had to restart it killing all my terminals, so nowdays I just run foot standalone, with sway tabbing and splits are kind of built into the wm anyways. But keep hearing about Ghostty and wondering if I am missing out on something.
The gem in this post is Pure, which I haven't heard of until now. I also have my prompt show the git status, and for large repos `git status` can take 10+ seconds to load and cache.
I had no idea that you could do that asynchronously, and then have ZSH update the already printed prompt with the status later! That blows my mind!
If you like that, you should check out my project https://beachcomber.sh . Its about time I take it from dogfooding to beta users if you want to give it a go.
"The single biggest win is what's not there: no oh-my-zsh, no prezto or plugin manager. I've honestly never understood the appeal of these frameworks."
"Most of these optimizations are about leaving stuff out. It's about being intentional and only adding things you're going to use."
I don't use X11 or a similar graphics layer, only textmode. Thus I don't use a terminal emulator
I don't use zsh. I use NetBSD sh
Smaller and faster
This is what I am comfortable with
Others may have their own preferences; to each their own
I might not understand others' preferences but that's their business, not mine
Great post! There are some neat tricks around completion initing that I'll have to grab. I use fish shell and have done a bunch of optimization around async git statuses too.
On Cygwin, FWIW, it pays huge dividends to avoid making the shell fork at all costs. Don't use $(sed ...). Use ${variable%foo%bar} or whatever. Cygwin punishes you hard for unnecessary fork().
As it turns out, avoiding unnecessary fork() is good hygiene everywhere.
In my life I can spare 50ms waiting for an terminal. But I have no time to spend 10000000 ms commuting to work, cleaning poop after an animal, or waiting for partner to put their face on!
The problem with this article is that the benchmark method they use is flawed. The documentation of zshbench explains why: https://github.com/romkatv/zsh-bench
Even with a low grade laptop, my zsh config grants me a sub 5ms prompt and a sub 1ms input lag, and that's far more important than the exit time.
I'm surprised people are still using nvm, considering it's impact on shell startup time.
I can't recommend switching to mise highly enough: https://mise.en.dev/
I had performance issues with Mise which I reported here https://github.com/jdx/mise/discussions/4821.
Unrelated to Mise but related to zsh, there's also https://github.com/jeffreytse/zsh-vi-mode/issues/316. I noticed this plugin was causing a lot of delay. Learned a decent amount about zsh profiling from that issue.
I used to use volta but then they killed it and told people to switch to mise. The mise setup is just way too complicated. I just am tired of having different config files for different tooling when it should just read whats in the package.json and be done with it.
I use aliases that lazily load nvm when I actually want to use it. This converts the shell startup penalty into a node/npm startup penalty.
This wasn't an option for us because as an org we used their recommended hook (1) to automatically change node versions when switching directories, but it effectively undoes the lazy loading.
With mise you get the behaviour of automatically switching when you change directories effectively for free.
1. https://github.com/nvm-sh/nvm#zsh
Edit: unless you aliased it to `node` or `npm`, which would be fine I guess but super annoying if you ran node or npm commands often. It is not worth the hassle, no one should use nvm in 2026 imo
How often do you launch a fresh terminal though? I start mine with a script to have favourite tabs ready at boot and then generally not much afterwards.
Constantly. I think we've used the excuse of "well, what if you just launch it less often?" enough to excuse bad performance defaults, especially when alternative solutions fix the issue with very few trade-offs.
Maybe once a minute, though it's bursty so more like 4 times every 4 minutes, but I still didn't notice the nvm slowdown
For every query and commands. I don’t use a DE, so pretty much everything is cli based. I use xterm and it’s bound to mod4+Return for me.
I wasn't actually aware of the impact. I measured the zsh startup time locally (with mvn active and commented out) and it indeed makes a difference (.39s -> .08s). Not that I would have noticed that without measuring :) - yes I'm an old geezer.
Thank you for the recommendation, I might then also be able to ditch sdkman as well.
With zsh i set up nvm to lazy load, so I don't pay for it when I don't use it (I'm a C++ and Rust dev, but I occasionally need to run js stuff from other team members).
I can strongly recommend lazy loading in zsh in general, I use it for pyenv too (which is also slow to load, but I write Python maybe every other week or so only).
The way to do this is to use the autoload functionality in zsh and have the autoloaded script replace itself with the real shell init code for the tool in question.
Am I the weird one? I usually have 3/4 terminals open at a time and rarely open new ones. Terminal startup speed is a non-issue for me.
The only thing I demand to be fast on my terminal is grep reverse search (ctrl+r) and of course typing a character. But if your terminal can't keep up with your typing speed there is something deeply wrong with it.
I open and close terminals _constantly_, but I'm usually pretty weird in my workflow, so no comment on your first question.
I run a scrolling WM and have settled into a habit of opening terminals when I need them, then closing them right after. I'll open a terminal, git pull, close it. Etc. I also use a terminal that launches cold in 10-20 ms, so it's not like a pay a price for it.
This is actually what I thought this post was about! But then I saw the Ghostty reference, which, in my experience, is not very fast to launch at all. I got it opening new windows quickly by running the main process as a systemd service, but Foot launches way faster without all that fuss, and allows you to go the daemon route if you want it _even_ faster.
EDIT: Just want to clarify, no shade on Ghostty. That project is cross platform and uses the 100% defensible decision to use the full GTK stack on linux. Foot is Linux AND Wayland only, and uses that very restrictive environment to optimize the hell out of startup and general performance.
I constantly open and close terminals too. Maybe I'm doing a quick lazygit check on cwd. Maybe I'm opening up an ephemeral claude/codex session for a couple questions about why a test failed. Or quickly editing a file with vim. Or remembering where I put that file with yazi or fzf. -- I don't even know, but all of it is contingent on it being fast to open a new terminal in cwd.
So much so that I vibe-coded my own terminal emulator for vertical tabs on macOS (using libghostty for the terminals) that is faster and less weird than iTerm.
+1 for Foot, truly a joy to use.
It can affect shell subprocess startup time as well, which, depending on your setup and the tools you use, might be worth optimising for.
I don't remember when I did it, but it looks like I must have gone through this at some point (maybe due to using GNU Make a lot? Or perhaps it was some other tool) - my zsh setup does a bunch of autocomplete setup only in the interactive case, and it seems to help a bit with startup time, at least on macOS:
For the interactive case, I don't really mind (within limits - the worst case, on macOS after a reboot, takes several seconds, and that's tedious). I also start new interactive terminal sessions fairly rarely.> But if your terminal can't keep up with your typing speed there is something deeply wrong with it. For the poor sools that had to work in VDI with radio link...
I open terminal far more often than that. But you should also remember that the startup cost is also paid by some subshells, and any shell scripts you run (the actual cost will vary: which init files are sourced varies between interactive and non-interactive shells as well as login shells and non-login shells, but it won't be zero cost).
Maybe, but notice that almost all of the article, and its followup posted today, is about the speed of starting a shell not of a terminal emulator window. So if you use interactive sub-shells, anything from shelling out of a vi clone or mailx, through :terminal in NeoVIM, to running tmux/screen or script, then the concerns are relevant.
That said, it is almost totally about the Z shell. So you might still qualify as 'weird' in this case on the alternative grounds of not using the Z shell. (-:
I'd say your workflow is pretty typical, from what I am seeing.
Developers that are very heavily invested in terminal and (over) optimize their terminal configuration are a small but very vocal minority.
I only notice how efficient my terminal is if something dumps a ton of logs to stdout. Sure I can pipe to a file and use vim, but it's convenient not to need to do that. And sometimes I have like 8 things going at once.
I do the same, so there's a good chance you're (we're) the weird one
Notice a lot of the article was about completions, which fish shell does better than zsh.
Speaking of slow, what I absolutely cannot comprehend is why ghostty is so popular. Despite being written in Zig it is very slow and a total CPU and memory hog. Just sitting there idle it’s pulling a constant 40% of my CPU? No thanks!
I'm the creator of Ghostty. This isn't right. It should idle at 0 to 1%, as supported by sibling comments. If you can collect more details about your system please open a discussion on the main Ghostty repo. Same with memory.
In terms of speed, same thing: if you can provide some kind of objectively measurable thing, we can look into it. Everything we've measured so far firmly places Ghostty in the "fast" camp (with friends such as Kitty).
We're sometimes faster, sometimes slower, but in any case not noticeably so. You wouldn't pick Ghostty vs Kitty for example for performance, it'd be something else. But you would pick Ghostty over say... iTerm2 for performance (but you may pick iTerm2 for features, its extremely feature rich!).
My ghostty is using 0.4$ - 2.5%?
> Just sitting there idle it’s pulling a constant 40% of my CPU? No thanks!
You need to figure out what is going on because that certainly isn't normal.
Literally seeing 0.0% on Linux
That happened to me too, turns out my desktop was software accelerated because I had screwed up the GPU config somehow. I asked Claude to fix it and it did.
I suspect you enabled some weird setting that you've forgotten about. ghostty isn't unimaginably fast but it's faster than iTerm 2 which is plenty. And I'm sitting here with a lengthy Claude Code session open, as well as a couple tabs for my docker container and dev servers, and its idle CPU usage is 0.0%.
WezTerm is great too and, AFAICT, is everything Ghostty wants to be. Kitty is also very good, although WezTerm edges it out for having an integrated muxer.
I can't understand Ghostty either except as some kind of trendy memetic thing, the $GME of the TTY world.
Yup. For a while I was switching between all three to see which I liked most, and I always end up coming back to Wezterm. The only issue I’ve encountered is that releases on various package managers are routinely broken, but building from source solved that.
This. Ghostty is significantly slower than Kitty on my Arch setup. It's still fast enough I guess, but no reason to switch from Kitty for me.
I know ghostty is designed to be super high performance but I find it both uglier and slower than Iterm2!
I started using it expecting to love it, but in reality it didn't seem to gain me anything but in fact was worse in several major ways. Also less configurable than Iterm2. :/
30ms is pretty slow
Life is too short to waste time on things that don't matter.
Maybe I should remove the following bit of code from my profile?
It's a whole chain of interpreters firing up (sub-shells, Perl for the cow, Ruby for the lol, I think.) :DBut what would life be without a little bit of fun?
"I tell you, we are here on Earth to fart around and don't let anybody tell you any different."
— Kurt Vonnegut
I know nobody is missing it, because it is the first bit of the blog post, but the author does have a follow-up where they note corrections based on push-back they received from a reader.
Apparently for some of the simplicity-produces-speed arguments, users have found complex/featurefull. tools that are still quick. I’m not sure how to evaluate this (I like simplicity just because it is easier to fit simple tools in my head) but we should note the counter argument (and applaud the follow-up).
I read Ghostty runs in a single process, but whenever I tried something like that eg a client/server model in urxvt or foot, I ended up reverting, because eventually some weird state affected the daemon and had to restart it killing all my terminals, so nowdays I just run foot standalone, with sway tabbing and splits are kind of built into the wm anyways. But keep hearing about Ghostty and wondering if I am missing out on something.
The gem in this post is Pure, which I haven't heard of until now. I also have my prompt show the git status, and for large repos `git status` can take 10+ seconds to load and cache.
I had no idea that you could do that asynchronously, and then have ZSH update the already printed prompt with the status later! That blows my mind!
If you like that, you should check out my project https://beachcomber.sh . Its about time I take it from dogfooding to beta users if you want to give it a go.
Follow up / errata post written by the author today: https://mijndertstuij.nl/posts/what-i-got-wrong-about-fast-t...
"The single biggest win is what's not there: no oh-my-zsh, no prezto or plugin manager. I've honestly never understood the appeal of these frameworks."
"Most of these optimizations are about leaving stuff out. It's about being intentional and only adding things you're going to use."
I don't use X11 or a similar graphics layer, only textmode. Thus I don't use a terminal emulator
I don't use zsh. I use NetBSD sh
Smaller and faster
This is what I am comfortable with
Others may have their own preferences; to each their own
I might not understand others' preferences but that's their business, not mine
> Thus I don't use a terminal emulator.
Yes you do. It's the one built into your operating system's kernel. You'll find a lot of, but not all of, its code in /usr/src/sys/dev/wscons .
Great post! There are some neat tricks around completion initing that I'll have to grab. I use fish shell and have done a bunch of optimization around async git statuses too.
On Cygwin, FWIW, it pays huge dividends to avoid making the shell fork at all costs. Don't use $(sed ...). Use ${variable%foo%bar} or whatever. Cygwin punishes you hard for unnecessary fork().
As it turns out, avoiding unnecessary fork() is good hygiene everywhere.
In my life I can spare 50ms waiting for an terminal. But I have no time to spend 10000000 ms commuting to work, cleaning poop after an animal, or waiting for partner to put their face on!