Telemetry is now optional in Warp(warp.dev) |
Telemetry is now optional in Warp(warp.dev) |
I don't want to stoke the flames again, but even VS Code doesn't force you to log-in. Microsoft knows that blocking you from using basic functionality of a free app is a bad user experience. It's good that you're letting people opt-in to a more private experience, but that was hardly the largest problem I could see the last time this was brought up.
From where I'm standing, it looks like you're trying to build a terminal for a pretty small audience; people who own a Mac, but don't want to use iTerm2 or the builtin terminal. That leaves a markedly tiny audience of people who don't use custom shells/incompatible configurations, want to log-in to your application and don't care about any of iTerm2's extra features. It's a bit of a pipe-dream to be honest, and you're not helping yourself by delaying your versions for other platforms.
My intention is not to discourage you, but I think this project needs a little tough love. You might be optimizing for the wrong users.
Re: login, I get the concern and we are exploring product options that let folks preview warp without login.
From a product perspective our goal is to make the terminal cloud-native and have a way of facilitating collaboration, and it's not really possible to build that without user identity. Specifically, login allows us to build cloud-oriented features that make the terminal have a concept of “your stuff” and “your team’s stuff” – for example Block Sharing. This is the same reason other collaborative apps like Figma and Github require login. We do get the concerns though and understand that this is not traditionally how a terminal has functioned and that it will make some users uncomfortable. But on the whole we feel like it's the right way to push the command line experience forward.
Re: configuration issues, we are trying as fast as we can to fix them. It's hard technically to both innovate on the command-line and maintain complete backwards compatibility, but that is our goal.
Re: Mac only, this is also really just a limitation of eng bandwidth, not a product strategy. We are 100% planning on bringing warp to more platforms as fast as we can.
> I feel pretty strongly that the terminal ought to be cloud-native
This seems like an oxymoron to me. A terminal is not cloud-native; it does network with the host machine, but a local shell is inherently local. If you've built the cloud part first, then you're building this product in reverse.
You're welcome to explore whatever avenues please you, but I'm not convinced there's much business to be found in cloud-native local shells.
> configuration issues, we are trying as fast as we can to fix them.
No sweat, I don't actually use Warp. This isn't a legitimate concern for me to level, just an example of the infinite treadmill of issues that you need to solve in a closed-source project. If alacritty has a configuration issue, the community fixes it. The way you've structured your project makes it extremely hard for your community to help you here.
> We are 100% planning on bringing warp to more platforms as fast as we can.
I wish you luck, it's hard work carving out a market segment of your own. I look forward to how your future features stack up against other terminals.
I vacillated for a long, long time before registering for an account. The distrust of cloud centralization, especially among developers who would otherwise be impressed by the “built in Rust” marketing, runs deep. Even if your intentions are totally beneficial right now, companies can change hands to people who aren’t aligned with user interest. Users who are developers are frequently in privileged positions at their workplaces, and being compromised at the level of their terminal could be catastrophic.
I’m the type of person who will take on many extra hours of configuration hell because I spurned Docker Desktop for its account requirement. So why did I end up signing up for a Warp account? Because yes, you are actually innovating in an area that desperately needs modernization. You recognize the risks and importance, at least in words, and in actions such as you have taken today. But because of the account requirement, I can’t let myself become locked in to Warp, so I’m not sure how deeply I’ll be able to explore it. And if a competitor arises that can innovate without any potential for lock in, I would jump ship without hesitation.
Thank you for your efforts and please, please, please take all the feedback here with utmost seriousness.
While it is possible to share a session with tmux, I have never used it for collaboration - in my 20 years in the industry, no such need has arisen. I'm not saying it's not a useful feature, others probably use it, but still, it's hard to see it as a selling point. And yes, mandatory registration discourages me from trying Warp.
It should be optional.
A lot of Mac users either don't know they're using zsh by default now or simply don't take advantage of it's scripting language/featureset. For those people, I think fish makes a compelling upgrade. No log-in, just a slightly nicer interactive terminal prompt.
Warp team: I watched the demo and it looks great, and I downloaded it while going through this thread. I didn't install. I think you're boxing yourselves out.
As a response to your questions, please see how we’re thinking about open source here: https://github.com/warpdotdev/Warp/discussions/400
As a side note, we are open sourcing our extension points as we go.
The community has already been contributing new themes (https://github.com/warpdotdev/themes). And we’ve just opened a repository for the community to contribute common useful commands. See: https://github.com/warpdotdev/workflows
The first thing I noticed is that it is fast. Noticeably faster than iTerm, which I use daily. Printing a large file doesn't stutter at all. As a test I printed (cat) a large video file. Warp finished in half a second, where I had to kill iTerm because I got bored watching all the garbled text fly by. Now, I'm not often printing video files to my terminal, but boy is it annoying when I do something dumb like that on accident. Sometimes iTerm just stops responding.
Other than that, there's some great feature innovations, like keeping the prompt visible even when scrolling or treating every command as a block. The block idea is nice because I can pin (bookmark, I guess) the output, so I can easily refer to it later instead of constantly scrolling around to find it. Maybe other terminals do this, but I've never seen it.
Also, aside from turning off telemetry, I didn't have to do any configuration to get it in a usable state. With iTerm I spend at least 30 minutes poking around the hundreds (thousands?) of settings to set everything up.
Now, there are some features than need a bit of work, like the AI stuff is pretty basic. It can't really do complex commands, like anything involving pipes or xargs. Tab completion is a bit clunky, like it will try and cd me into a text file.
Anyway, it's clear to me now that I've become complacent with iTerm and there's still a lot of innovation on the table for terminals. I think a new generation of developers that care less about open source and privacy will like this. However, I'm mostly happy that there seems to be real innovation here. Maybe we'll see some of the best ideas make their way into other products?
There is a lot of negative thoughts on this thread - would love to hear from the people who actually have tried out Warp and not those waxing on about the indignity of being asked to login to a free product.
Warp is built on alacritty which is firmly middle of the road in that chart.
As a side note, we are open sourcing our extension points as we go.
The community has already been contributing new themes (https://github.com/warpdotdev/themes). And we’ve just opened a repository for the community to contribute common useful commands. See: https://github.com/warpdotdev/workflows
Hope you find this helpful!
- If you want fuzzy command, history, file / contents search, use fzf [0] (you should probably be using fzf and ripgrep [1] anyway if you work daily in your terminal).
- If you want sessions, use a multiplexer like tmux [2] or zellij [3].
- If you need to have your own "cheatsheets" use navi [4]. If you want to sync them with your team, use whatever sync solution you like.
- If you think you need a text editor in your shell's command line, reconsider. If you *really* want to edit and re-execute the last command in your editor of choice, use something like "fc $!" [5] or create your own shell solution for it.
- If you want a sexy prompt use starship [6].
- If you want terminal sharing use tty-share [7].
- If you want to ask GPT for help, don't do it in your terminal. Open up ChatGPT (or whatever future UI will exist), ask your question, and check that there's nothing harmful in what your about to execute. Sometimes friction is good.
If you want a terminal that's pretty out-of-the-box, where things are "clickable", you don't have the time or interest to invest your energy in learning tools that could massively boost your productivity for years to come, and you don't care about designing your own workflow and being able swap parts of it at any point, without depending on any single "app", and if you don't mind "logging into your terminal" (what the actual fuck, excuse the language) or the terminal adding its own SSH wrapper and doing things you don't know to the hosts you connect, then maybe Warp is OK for you. But then again, maybe you're not going in the right direction.
There are so many more awesome ways you could improve your shell experience than making things clicky. I don't understand what the market is for Warp, is it for wanna-be professionals that can't be bothered to become professionals? I completely fail to see how this could succeed as a paid product, especially with a subscription model.
[0] https://github.com/junegunn/fzf
[1] https://github.com/BurntSushi/ripgrep
[2] https://github.com/tmux/tmux
[3] https://github.com/zellij-org/zellij
[4] https://github.com/denisidoro/navi
[5] https://shapeshed.com/unix-fc
[6] https://starship.rs/
[7] https://github.com/elisescu/tty-shareHave enough developers really been conditioned to accept surveillance and pervasive centralized control that someone thought this was a good idea? The fact that the telemetry code even exists is a problem - both in case of a bug that turns it on, and what it says about the proclivities of its developers.
It's utterly inappropriate to refer to this as "the Rust-based terminal", as if the language is its distinguishing feature. More like "the obscene SaaSification of a basic foundational utility". I'd be very much interested in trying a new terminal written in Rust, which could presumably be quicker and have a smaller chance of hiding latent parsing bugs, but this most certainly is not that.
Easily share out custom workflows to fetch development keys and login to different services. Onboarding scripts that setup all the terminal things needed for development environment.
> "the obscene SaaSification of a basic foundational utility"
Watching their demo vid, it is a very nice terminal that also supports org wide shared scripts, in a unified manner with autocomplete.
Right now I have a text file with commands that I copy and paste daily to get things done. I share bits and pieces of this file with new hires when they onboard to solve various common problems. This shell solves that problem for an entire organization, and it does so in a discoverable way.
Can an org have their own repo of shell scripts? Sure, but then you have to make sure all developers in the org pull the latest when changes happen, and discoverability within "folder full of scripts" isn't the best.
> The fact that the telemetry code even exists is a problem - both in case of a bug that turns it on, and what it says about the proclivities of its developers.
Telemetry exists because running user studies is expensive. "What features are people using, what is crashing, where are people getting frustrated".
Hacker culture is very minimal these days.
Now a quick npm install here and there, using a slow af editor like vs code, using autocomplete which sends your strokes to the sky and using every single cloud service available to be "pragmatic" while sipping soy lattes, have made this whole industry, slow, uncool and boring.
Seriously, what's wrong with you, iTerm2? Feel free to keep your multilevel settings, but expose the important stuff up front as a single click preset to apply.
> Juicero was an American company that designed and manufactured the Juicero Press, a fruit and vegetable juicer. The Juicero Press was a Wi-Fi connected juicer that used proprietary, single-serving packets of pre-chopped fruits and vegetables that were sold exclusively by the company by subscription. From 2014 to 2017, the San Francisco-based firm received $120 million in startup venture capital from investors.[1]
> The company attracted significant negative media attention when consumers and journalists discovered that its juice packets could be squeezed just as easily by hand as by the company's expensive machine. On September 1, 2017, the company announced that it was suspending sales of the juicer and the packets, repurchasing the juicer from its customers and searching for a buyer for the company and its intellectual property.[2][3] After its collapse, the company was described in the press as a symbol of a dysfunctional Silicon Valley culture. The Guardian wrote that Juicero was an example of "the absurd Silicon Valley startup industry that raises huge sums of money for solutions to non-problems."
https://en.wikipedia.org/wiki/Juicero
I laugh, but at the same time, I have a Nespresso machine.
Conclusion: Subscriptions work better if you sell drugs.
I tried Warp and I wasn’t impressed at how it outputs text. I had some issues copying and pasting because it formats some things weird.
I mean, it's ok? But I rarely feel like opening it.
Small things don't work for some reason. Like alt-clicking in nano doesn't move the mouse like it does in basic old Terminal.
And the tab completion is really aggravating for me. There's like 3 different actions it takes almost randomly that make zero sense, for example in a window I had open:
- If I type "na" and press tab it shows me a popup window with suggestions.
- If I type "nano", it shows the entire command I wanted to run based on history "nano test.log"... but pressing tab does nothing. You're supposed to know to switch to the right arrow.
- If I type "nano test.lo" there's only one possible completion and now tab does work like the right arrow
- But hold on! If got a selection window in an earlier step, and I decide to type to narrow down the selection so there's only one option left... tab doesn't work like the right arrow and will instead stupidly try to cycle the selection window even though only one option shown is still valid.
It's like the selection window ignores everything you typed after it opened, which plain old zsh solved, but they somehow unsolved
Have also had it break for me once and I'm not clear on how to reproduce, it involved SSH but somehow the order of "blocks" for commands went out of whack, so there was a block stuck underneath the "bottom" where I'm supposed to type, and it'd steal input from the prompt.
And there's this really minor but annoying thing where if you try and select some text then right click to copy, the entire command block gets highlighted, even though in order to select a piece of text you had to intentionally click twice to unhighlight the command block.
-
Having little moments like that just introduce enough aggravation that I don't want to bother.
To me iTerm is a happy middle ground. It's barely any different from Terminal, but on the few occasions where it adds some value (like being able to view a high quality image preview in your terminal, or share commands between SSH sessions) its completely unintrusive, the opposite of what Warp seems to be aiming for.
-
Also, semi-off topic, but after using it I get seriously stressed out about the fact that companies like this are hiring people who need to make a living.
There's no reality in which this actually becomes something to justify taking in tens of millions of dollars in funding, which are probably expecting 100s of millions in returns. Clearly their plan is that some companies will end up having this be the go-to tool for developers, but the most "hardcore" developers are the ones who would be most repulsed by this.
The more casual developers work at places that see developers as a cost center and will never spend real money on a terminal when they're barely ok with paying for a Keurig.
I imagine that the founder and maybe the core team know they can get cushy jobs with their technical skills and resumes so it doesn't bother them, but the problem is these things tend to scale up before falling apart.
Eventually they start to pull in people who don't have those kinds of luxuries just to function. People who didn't go to some top school or have a string of in-demand logos on their resume might not have what it takes to weather this all falling apart like they do.
Like yes, startups are inherently very risky, but again, the further away from the core team you get, the higher the headcount they have to show before hiring people, the longer they've been around, they'll start to attract people who can't reasonably be expected to think they're being hired onto the Titanic as it leaves port.
Coincidentally, Warp is based on Alacritty
They had CNC machined parts when they had the capital to injection mold everything (they could have used exotic alloys and it still would have been cheaper), complex multi-material plastic parts, and the whole thing was way too complicated - the door lock alone had dozens of parts in an assembly controlled by a dedicated PCB.
If software developers had a professional oath to do no harm akin to doctors, creating such systems would be professional malpractice. Some lines just shouldn't be crossed.
akin to engineers as well.
Weapons need software, too.
A $20M Series A requires those investors to believe the startup can get to $1B for the math to work out for a typical fund.
Nobody has ever defined or proposed defining "expect" as "I can see how this could happen" until just now. Not only does "expect" imply 50%+ odds, it implies something like 90%+ odds. It implies that you would be surprised if it didn't happen (because you were expecting it to happen).
The target market is developers, and developers are spooked out by this kind of thing. Seems risky.