Warp terminal – no more login required
warp.dev
warp.dev
I think I'll just stick to iTerm.
Warp is just something that will, quite honestly, never appeal to me. I don't think the UI is bad, I don't think their featureset is bad, but I also have no desire to use a proprietary terminal app. Tilix is open source and has worked fine for the past 10 years - I'm not giving it up for a chic frosted-glass background and some syntactic eye-candy for bash.
My fear of paid software mostly stems from not actually owning what I pay for. I know people who are (horrifyingly) unable to use Git without a paid GUI that handles everything for them. If one day they update their computer and it's unsupported, or if the developer quits updating it, they've gotta learn a new UI. I'd rather just learn how to use the original tool quickly, and save myself a few hundred hours with 2 or 3 macros.
To each their own. I've paid for too much software that has gone to shit in the past, and I'm perfectly productive with free tools I can actually own.
Genuinely curious because I feel like I'm missing something. I tried it again a while ago and still couldn't see anything I couldn't already do.
- Multiple profiles
- Logging of terminal output to files
- Regexp search
- Search marks matches in the scrollbar
- Quick jump between prompts — it knows where prompts are in the output, so you can quickly scroll to each position
- Window splitting
- Filtering — like find, but filters the window to only show matches
- Triggers: Do an action such as bouncing the app icon or playing a sound when a text string occurs)
- Status bar: Show the Git branch, current directory, etc. — better than custom prompts
- Preserves windows/tabs on restart/reboot, even current directory and window contents
- Cmd-Click to open files and URLs
I bet Terminal.app has gained some of these features over the years , but it certainly doesn't have have all of them.
- Minimal theme mode for a more simplified window appearance.
But yes you're right, most of these are in the Terminal+Tmux combo already.
Not sure a lot of people would agree. Tmux is way less intuitive than tabs/split panes directly supported by the terminal emulator. Also tmux is dog slow compared to native tabs.
split pane vertical: cmd + D
split pane horizontal: cmd + shift + D
focus pane left: cmd + [
focus pane right: cmd + ]
Just gold standard imo. tmux is great software and very useful on remote machines but its keybinds don't come close.
Tmux also comes with some limitations that make it annoying (to me) to use 100% of the time :
- scroll back is painful compared to native terminal emulator scrollback, so is mouse support
- rendering gets slowed down by tmux because it parses the output itself, the fastest terminal emulator with tmux will feel slower than iTerm without tmux
- some TUIs work badly in tmux, neovim for instance. I’ve also had issues with htop in the past
Tmux is a great tool, but I can’t conceive using it by default, it’s too janky for that.
Already a deal breaker.
[0]: https://babbagefiles.xyz/quake-drop-down-terminal-history/
[1]: https://steamuserimages-a.akamaihd.net/ugc/18380429971797580...
just to clarify though, the point of the login is that we have features that cost money to provide like AI and collaboration, not anything more nefarious, but i get that it's a new behavior and reasonable devs might not like it.
They have signalled to me, loud and clear, that their product is not for people like me, and I will never try it. Doesn't matter that they've removed the requirement, we've all seen what type of product this is.
there is a large segment of developers who find the command line hard to use or who just want a better, more productive experience using it. to be clear, you may not fall into that bucket, and that's OK.
the point of the login is that we have features that cost us money to provide like AI, and we need some concept of identity to prevent their abuse. i don't think that's detestable (e.g. it's very similar to cursor or copilot), but i get that's a new behavior in the terminal and am sorry it put you off.
Deal breaker: I do not like normalizing the deviancy of sending shell input to a third party. F opt-out telemetry and a "trust us" privacy policy. They're pressured to deliver maximum profits in a short time, and having to use the internet to use your shell and paying for the privilege of handing over private work is absolutely absurd. It could all, and will likely, as most startups go, shutdown at some point in the future and then users will likely be left with nothing. It's a signal there's too much VC money available.
Fundamental error: It's an attempt to innovate in a crowded category inappropriately, like trying to make a physical proxy "phone" or security device that could've been an app, it's too tied to an app. It could've been reduced to a multi-shell, shell-oriented hook that would've been inherently multi-platform without the need for yet another desktop app and would've been portable to other environments without the effort or expense.
I would expect a technical user would be immediately skeptical at the value proposition. I can't even trust git to not break itself when I'm using it directly. Throwing in an LLM and natural language sounds like an express elevator to hell.
they did a reasonable job of integrating it with the terminal, of giving you explanations and clear points of review before taking any specific actions on your system, and in general i reasonably trust it won't do anything i don't expect or modify anything critical without asking first - that's been my experience thus far (no guarantees, it's an LLM after all)
for my entire career i've never been much of a bash person, i learn just enough to slice up some text file or something as needed, or put together a build process, and then it all leaves my brain because it's just not something i'm interested in. i used to lean on the tldr manpages pretty extensively when i needed to do a bunch of bulk file processing in the terminal or something.
the case where warp shines is the case where i'd have a folder of text files, or some data i wanted to parse and process real fast, and i'd be thinking "god, i could pull out my little bash reference, or i could throw together a quick python script to get this data in the state i need." i'd say it has an 85% first-time hit rate as far as "i need to modify this file in this way" or "i'd like you to find all the html tags in it with this attribute and extract the values into a list in a text file" or "can you convert all the HTML lists in this document to markdown lists" or "i just downloaded a huge CSV of logs can you just return the count of a certain token on each line in a newline-separated text file." just fast little things i'd either build a utility for, install a plugin for, or poke around online for a website to quickly do.
it's saved me a LOT of time in those cases. i went from only opening it when i wanted to use it for one of those cases to using it full time over the course of a few months and i am cautiously optimistic about it. i was not thrilled about the login requirement, and i'm glad they dropped it.
it also enables justified laziness - i'm doing something in ruby and i haven't used it in a long time, i'm just popping in for a minute - what command do i type to activate a virtual environment? i already know what i want to do, i just don't know the specific incantation off the top of my head. i can just type what i intend to do, and it reminds me of the answer and does it. easy. it also detected which virtual environment tool i'd installed years ago for me, so i got the correct one on the first try.
i obviously can't say whether "optimizing annoying one-off text-munging tasks" is worth the various tradeoffs and sharing data with the company, but it's the only LLM tool i've kept in my workflow after trialing it, and applied specifically to the cases it shines at it does a good enough job to be a time-saver (not my experience with copilot, cursor, so on.) i also wouldn't trust its output in a build process or anything that would be frequently run or impact production data. those are all strikes against it, to be fair.
just my two cents. i'm not gonna hard sell you on this at all - i'm not that passionate about it and only use the free tier - but it's probably the best case i've seen for "LLMs can help you in your job" out of every case i've seen.
The main thing that's a big of a pain is you'll probably want to set up a scrollback pager (I use neovim as mine but vim works too) so you easily search the terminal output and copy/paste from it.
I use
# https://github.com/kovidgoyal/kitty/issues/719#issuecomment-...
scrollback_pager nvim -u ~/.config/kitty/kitty-scrollback-pager.vimrc -c "silent write! /tmp/kitty_scrollback_buffer | te cat /tmp/kitty_scrollback_buffer -"
and the scrollback pager vimrc:
``` set nonumber nolist showtabline=0 foldcolumn=0" set clipboard+=unnamedplus autocmd TermOpen * normal G map # ^ map q :qa!<CR> "silent write! /tmp/kitty_scrollback_buffer | te cat /tmp/kitty_scrollback_buffer - ```
I looked at alacritty but I really like using terminal tabs and the alacritty dev is really really against them and I found the dev's attitude to be more than a bit abrasive.
Not that Kovid (kitty's dev) does not come off the same way sometimes, but I tend to agree with Kovid a lot more.
Does this help at all?
I was looking for a way to get a user prompt within wez and came across this page which has an example for what you asked for "Example of interactively renaming the current tab".
Other terminals are free, open, and more than good enough, why would I use Warp? I’ll pay for good development tools, it’s not about that at all, it’s just that Warp is a… terminal. I pay for JetBrains because it’s that much better than the free options (my opinion). Is warp that much better than iterm?
I do have an open mind, convince me that warp is worth it, that it is irreplaceable. Convince that I need it even though it will be enshitified like Postman.
Warp is good, it’s a step up from iterm for people who spend a bunch of time in the terminal but don’t want to invest their brain power into emacs or vim or whatever is real nerd software nowadays. $25/month is basically free for the amount of time I use it. I’d pay $25/month for anything that makes my life marginally easier.
Am I going insane? Why would you pay for this unless you have a workplace humiliation fetish?
I don’t care if a tool is free or paid, I care if it is going to make my life better. You can sit in the office laughing at me for paying to use software, I’m not bothered, because I won’t be in the office to hear the laughter, as I’ll be using all the time I’ve saved to do something I want to do.
Pay me $250k/year to write code and I’ll spend $5k of that on a paid for editor and a paid for terminal and a paid for git client and a paid for database client and a paid for rest client and a paid for shortcut app so that I can go home at 3pm every day and spend my remaining $245k on the things I care about instead of being stuck in the office micromanaging my vim configuration so that the nerds will respect me.
From the absurdly CPU-intensive landing page it seems the novelty is mainly some network storage and an LLM-integration, things I already have several of in my terminals on my work computer. I switch over to them in the blink of an eye because my window manager is extremely responsive and keeping their output localised and not split up in a single pseudo-conversation in my current main terminal is a perk, not a bug.
> We are planning to first open-source our Rust UI framework, and then parts and potentially all of our client codebase. The server portion of Warp will remain closed-source for now.
https://github.com/warpdotdev/Warp/discussions/400
None of the server bits will be open, ever, but some of the UI code will be? Better than nothing.
On a more serious tone, how could this thing raise 73 million USD? I wish one could short startups.
Edit: Wow, it gets better.
Their terminal only works on Mac, do they need to raise more $$$ to bring it to Linux/Windows?
Also, at $15/mo, they'd need about 400,000 paying users to even recoup the investment. Will that many OS X users pay $180/year for something that is essentially free?
How does this make sense? Did they accidentally stumble into AGI and are keeping this a secret except for their investors? Are they laundering money? This whole scheme makes WeWork seem sane.
Hypothetically if an LSP-server existed for command line args, it could be integrated at the shell prompt to create an autocomplete popover. Even cooler, generating LSP syntax should be doable en masse as an LLM can read every single man page in one pass and add what it learns to the supported syntax.
And for generating entire commands it's possible to create a desktop app: I'm imagining it having a text field to type what you need ("a command to double the speed of a video") and then it'd:
1. generate the command ("ffmpeg -i input.mp4 -vf "setpts=2*PTS" -c:v libx264 output.mp4"
2. copy the command to clipboard
3. display an explanation of each part of the command and what it does (raw markdown from the model should be fine)
4 (optional togglable preference setting) after paste is detected, restore previous clipboard content
This app could be created really fast now LLMs can code for us. And the command line generation could be done fully locally with ollama.
Switching back to iTerm2 again because of https://github.com/warpdotdev/Warp/issues/1811
I like the sound of the other two features (rich text editing and automatic AI commands) but the muscle-memory cost is too high.
Since I'm not the target audience I can't say whether this is both "automatic" and the thing you're looking for, but: https://iterm2.com/documentation-preferences-general.html#:~...
No ETA for this yet, but this is very much on our radar to fix asap.
It might be worth updating the GitHub Issue: the impression I got from looking at it was that it wasn't a priority (the most recent official comment is over a year old and comes across as quite dismissive of the request).
I don't think I'm much more productive at Terminal.app in 2024 than I was in 386 Linux console-connected to other computer via 14400 dial-up in 1999.
https://github.com/yonatan8070/dotfiles/blob/master/fish/.co...
I honestly don't use many of their features besides being able to ask the LLM for how to do something random in git and getting the answer without leaving the terminal. The way they organize each command+response into blocks is also cool for readability and copy/pasting.
Can I (yet) just press tab and have it inspect my PATH variable for completable items?
That is literally the only reason I gave up on it.
Good luck to the team, not for me.
[1] waveterm.dev
Those people don't use a terminal.
And even if config files were involved, do you think people configure their terminals (editors, whatever) not to increase productivity? Do you think we use vi as a sort of fidget spinner, and just edit our configs for the sake of editing them?
The other commenters are right: Warp has no real target user base. This, I believe is by design because their business model is Juicero for bash. They exist to farm VC money.
Also let's be real. All the default terminals are great and fast enough for typical use cases, the only ones that really care about GPU accelerated terminals are the ones that use editors in the terminal, so they're going to be customizing everything anyway.
Not because they aren't attracting enough users or something. No you see because that might require you to be at least a little wrong. So that can't be true.
Also most people have really good metrics and telemetry to measure their productivity and are really good at telling what makes them productive and how to invest in it.
No you see, don't worry, it is HN that is out of touch, not you.