Warp is now open-source(warp.dev) |
Warp is now open-source(warp.dev) |
for zsh:
autoload edit-command-line
zle -N edit-command-line
bindkey '^X^E' edit-command-lineWhat you describe sorta works, but you lose things like file/dir-based autocomplete, since your editor doesn't know about your shell session.
Alt-e for fish
Ctrl-g for Claude code
EDIT: well looks like this is not OS/2 Warp. I wish the title would have noted this is somekind of app instead of just saying "warp".
Nice...
All six users will be disappointed.
... it's not about OS/2 3 or 4, but some AI thing, and apparently feels no need to disambiguate or even acknowledge the prior usage of its name
... closes tab
I personally feel like there's no need for acknowledgement at this point, but that's just me.
Vercel Security Checkpoint
Failed to verify your browser
Code 99
Appreciate the candid take. Warp is great.
Warp is the only closed source terminal product I know of. Most other popular terminal emulators are open source already.
I feel like their funding is drying up and this is their last ditch effort to have the "community" build their product for them.
They claim agents will run the show, with inputs from community in the form of ideas/specs/direction. I wonder how long that will be sustainable for given the subsidized model prices are collapsing as we speak.
Is this an attempt to pivot to something else while the "community" keeps their first product alive? Maybe I'm being too cynical here, but I don't see this as an act of good faith, especially given their roots in VC funding.
Is Warp a terminal? Or an agent harness? Or both?
Warp as a terminal to me seems less interesting than having a well built agent harness like OpenCode that can effectively use many different models. If it's both, is there any advantage to having them be the same thing? Like, is there any way your harness can be smarter if it is also tightly integrated in your terminal? Or is it just something that Warp happens to do both of?
Agentic AI, broadly speaking. Including CLI agents and IDEs.
Sorry but I can't read this statement in a tone is not absolutely dry sarcasm.
Warp is a masterclass in how to have a great product statement, very nice UI and completely ruin it with greed, closed-source, and by not listening to your customers.
I don't know who at Warp got replaced and wants to "fix" things, but they have a lot to overcome. At this point, it might not even matter-- another product written from scratch might have better success.
Warp had an account requirement in the beginning which spoke volumes about the misalignment of values. Now the terminal is not called a terminal, it is "the agentic development environment" (whatever that means) which also lowkey implies that it might have some kind of online features. But at the same time I understand that it is now an absolute requirement to mention AI on any web page for any product.
Does it call home?
Very poor first time user experience.
Could we have just released it? Absolutely. But I think everyone who contributed felt better knowing that what was released had one final "ready for public" review. Then our regular review process handled that going forward.
Considering it came out in 2020 - a few years before the LLM hype train left the station, and when I started using it there was no AI integration, this doesn't seem accurate.
I liked using it because of the text input.
tangential: I've seen Mitchel tweet that people in SF have ran up to him showing him how they fully riced their Ghostty setup. How many people here have done this and how easy/manageable is it? e.g. just forking the repo and implementing whatever Warp feature I like?
Ghostty remains incredible stable and usable and fast compared to competition.
And the motivation was warp is doing a little bit more than a terminal.
Glad to see now warp is open-sourced
- Being able to drop in a brand new laptop and have almost all I need out of warp and a barebones ZSH configuration. I used to spend a lot of time on getting ZSH both fast and feature-rich, and it's brittle. Warp by default provides good auto-complete etc.
- Fast rendering and sane graphical defaults. I don't need to do much more than put my font of choice.
- QoL features regarding file rendering etc.
I've never used the agentic parts of it, when I needed CLI my company paid for Claude and I get most of my stuff done with what I get out of my Zed subscription. But I'd be more inclined to do so now.
That would be awesome!
The AI integration only came years later, and they probably figured maybe they could compete in the space with Cursor, repl.it, etc. And then Claude Code came and just devoured everything, while Google bought Windsurf, Microsoft pushes Copilot, and OpenAI has Codex (while at the same time kitty and ghostty also built really nice, fast terminal emulators)
I agree I don't think they have any real shot and I certainly wouldn't recommend investing in their next round, but it's not like this was their plan all along, they went where the winds were blowing.
I'm actually pretty proud of the final setup I've created with it.
Each time I start to implement a new ticket, superset will pull the ticket from linear, create a worktree/workspace, reserve ports, start the servers, start a browser and start Claude with the ticket as instructions.
The cool thing with this setup is, I can have like 10x the same servers running on different ports/worktrees. Each time an agent is done, I switch to the workspace, look at the browser and can immediately test things.
It's like having 10 virtual desktops. Wonderful!
I am skeptical that they decided human input is their bottleneck just as the cost per token spiked from some AI providers. I see this as a way to reduce their compute spend (offloaded to the community), but I doubt they are going to give up any creative control, so their employee review bottleneck probably won’t change.
edit: nevermind, it was quite discoverable...
As they've been scrambling to find a way to monetize and riding the AI train, it feels more bloated than ever and the constant pushing for me to use "agents" and whatnot really put my off using it. Plus with all the privacy concerns I can't with good conscience use it on my work machine.
So yes, I'd like a non-tracking, no-AI version of Warp too.
I don’t think the approach of open source as a substitute for a quality program is going to last.
i currently use tmux and ghostty for my workflow
Absolutely, in heaps. The very moment you start Warp, it sends 5 HTTP requests, (1 version check, 1 LLM model list, 3 telemetry events). The former two go to app.warp.dev, the latter 3 go to warpianwzlfqdq.dataplane.rudderstack.com and include a persistent UUID, your operating system and version, Warp version, and the tracked event name and its properties.
All of this happens before you even see the window. After you're done clicking 'No' to all the SaaS nags, you can turn off telemetry in the settings, but for some reason it gets turned back on when you restart the terminal.
While the terminal is running, it calls home whenever you trip one of the events listed in this 7000-line long file: https://github.com/warpdotdev/warp/blob/d0f045c01bacbd845a63.... Besides all the hosts listed earlier, it also makes requests to o540343.ingest.sentry.io.
All in all, a privacy nightmare.
To add more color to some of your points in the post above:
* We give users the option to disable telemetry before sending any back to our servers.
* We use Sentry for crash reporting, this can be disabled.
* Yes, we record telemetry with the option to disable. The events are in our source code (as you point out) and also on our website.
* We have a network log in app that you can open and view _any_ request we make, including requests to send telemetry.
* If you compile the OSS build it has no telemetry or crash reporting.
Our motivation with open sourcing is to build more trust with our community, not less. Happy to discuss more and, of course, the codebase is available to audit.
At this point I'm not sure we share the same understanding of what constitutes trust.
To me any connection initiated by a terminal app itself is terminal (pun intended) for its reputation, leave alone hideous things like online login or telemetry.
But that's just me: I'm old fashioned. And I'm sure plenty of people will continue to enjoy Warp.
Do you regret having this requirement in the first place?
Personal feedback: I live in a terminal 24x7 for the last 30+ years and once Warp came out I wanted to try it out immediately, but I was impressed by the requirement. So I never had a chance to try it out.
This is a terminal that is not designed for users.
> not designed for users
It's not designed for you.
I think there's a reason people are very interested in libghostty. It's a fully-featured library to build off of, but Ghostty proper still lacks a lot of polish.
I welcome the experimentation, there will definitely be something new, but this ain't it. New primitives are needed, at a higher level of conceptualization, not merely a fancy new interface.
> writing code is now fast, it's getting cheaper, and quality is going up to the right
My hobby projects have 100x more tests than they used to, because LLMs are great at writing tests. And my subjective experience is that the net quality has increased as a result.
YMMV, but it’s certainly a common belief, and for me at least a lived experience.
I don't care what people believe, I care about what is. What is measurable. What is factual. I need evidence for a belief to be meaningful. I need strong evidence for a belief to be strong. Not just evidence in favor, but evidence that alternative explanations are unlikely.
Currently I see evidence that things are moving fast. But I am unable to distinguish if this is actually because of AI or because increased efforts and motivation. Most importantly, speed isn't the same thing as velocity.
What I do not see evidence for is increasing quality. In fact, I see strong evidence to the contrary. I see strong evidence that quality is declining even quicker than it was before. I'm not convinced AI is that cause of this, but there's more than adequate evidence for me to believe it is a (significant) catalyst.
Right now you could throw a stone in a random direction and there's a good chance that whatever it lands on will be decreasing in quality. It is even easy to gesture broadly at Microsoft, but they aren't the only big tech disappointing their users.
I hear a lot of claims. I don't see a lot of evidence.
For my part, I’m too busy building to spend time quantifying quality issues in my past decade of hobby projects so I can compare to my new work, which subjectively is far more capable and higher quality.