Ladybird has been really shaping up over the last month or so, I pull and build the latest bits every morning and use it for low stakes web browsing (HN, news sites, reddit, youtube, blogs, etc.) and it's getting very close to the point where I forget that it's a pre-alpha.
Very exciting to watch a completely new browser/browser engine come together. Reminds me of the early days of Mozilla and Gecko
I suspect in a year or so, an entire browser can be prompted by AI and then it won’t feel all that cool anymore. Unless AI hits a wall soon (unlikely).
I've actually used AI to create a WebKit port (not an original browser). The amazing thing is that it worked--but it still took months of issue triage. Even when the AI was working on its own, it had to spend enormous amounts of time just on compiling and testing. There are so many moving pieces, and each individual issue could mean the AI spinning for days.
Even if a future AI can drive web browser development completely autonomously, I suspect it will take a huge amount of real-world time because that's just the shape of the work no matter how "smart" you are.
Yep, I've been building a custom H2/H3 server for nearly 6 months now maxing my Claude 200 plan every week. Very pleased with the results but it's definitely not a quick or cheap task, it's constant decision fatigue and tons of rigorous testing on every piece in turn. Fuzzing, mutation testing, Miri, protocol conformance suites, H2load, actual networked runs, so on and so on.
I mean it worked, if scaling doesn’t stop and it can do this non stop at GPT Ultra speeds, this whole task will be done in weeks, maybe lesser. Doesn’t refute my point, ladybird is unfortunately going to be a pointless project soon. Though I guess if you have fun, who cares.
No, because the bottleneck isn't just model speed. The model has to literally compile the engine and run the massive test suites. Some of that could be sped up if you had a real build farm instead of my little computer, but there's a limit to how much you can parallelize things. And remember, I was porting WebKit, not remaking it, which would be orders of magnitude more work.
So you're not going to get to "build me an original web browser from scratch, make no mistakes" and get a new app the next morning or even next week, I don't think.
I'm doing a web browser from scratch, no web or js engine reused (only general-purpose libraries allowed), in my spare time/tokens on a Claude $100 sub.
After a month or so, many simple sites work, and now the bottleneck is the stupid amount of JS sites use for everything, and Cloudflare which doesn't like someone using a non-officially-approved browser.
Absolutely not a replacement for Chrome or Ladybird, but it can be done already.
"[...] as far as we know, no other browser can do this today. Here’s a text animation while JavaScript blocks the main thread in 1.5 second bursts, with Ladybird in the top left:"
This is amazing!! I'm really surprised no other browser is able to do this. Is there an implicit "security" issue with it, or maybe standardization when it comes to keeping some values from DOM-Javascript in sync? Or simply an oversight in their part? In any case, congratulations to Ladybird's team. Computers are so much more capable of raw performance and computing, and it's nice that some developers out there do care heavily about it
afaik the decision to run animations on or off the main thread is a recurring point of debate not just in web browsers.
On iOS for example SwiftUI runs them on the main thread while UIKit w/ CoreAnimation runs them off it. I don’t remember the specific reasons for these decisions but the point is, as with many things there are pros and cons to both approaches.
Is it just me, or Google Maps is really broken in Ladybird?
I remember the first time I compiled it (IIRC in June), I was blown away, it was a bit slow, but it worked. (My standard browser performance test is Google Maps - zooming, panning, street view etc really shows the browser's capabilities.)
I just compiled the latest, and Google Maps is almost completely broken. You can kind of move around and zoom, but almost all the widgets are missing. The search bar is not missing, but it doesn't do anything.
I get that this project is still in development, but given the progress in the past few months, it's just strange that one of the most important pages on the internet is so disfunctional.
ps. Bing maps works much, much better. Is Google behind this?
Honest question for people. I'm a C++ dev and I'm seeing a lot of moves to rust. Everyone is using agents to move stuff to rust and I tried it myself. The amount of insane hardrive space usage (over 90GB of build artifacts that constantly need cleanup) is that normal? What am I doing wrong. Are people seeing advantages after a move to rust that wasn't there in C++ (assuming no new features). Like is your project better after moving to rust in obvious ways?
I have the same problem and am regularly `cargo clean`ing hundreds of gigabytes of build artifacts. It is a real problem. I feel like I need minimum 4TB of SSD space just to comfortably work at the scale I'm working now in Rust. That's just for the working space of the target folders across worktrees.
I just think agents benefit a lot from having unavoidable feedback in the dev loop, and Rust affords that through its strictness. It's easier for agents to fall into the pit of success with Rust. I do think it's somewhat lacking in compile-time features, though.
It's expected. Cargo has very conservative cache cleanup (once per month IIRC).
There's incremental compilation cache that generates lots of churn (although Rust's compilation unit is a crate, "queries" in the compiler are cached individually, creating many cache entries per function per type).
The other half of the disk space is debug info, because abstractions that optimize down to zero runtime code still carry all of their source-level debug info.
These can be adjusted in build profiles if you're ok with coarser debug info or less fine-grained caching.
Shared libraries (i.e. not building every damn dependency) would fix this problem real quick in many situations, but of course, Rust people don't want to commit to an ABI in order to keep their options open. I don't think it makes sense much longer - at this point "can't do shared libraries" seems more like cultural inertia.
Two or so ABI breaks with a few years between them also seem quite survivable.
it was days, but it was related to something the agent was doing with tests and targets, where it was basically compiling a billion variations of the almost 500,000loc, 1100 deps, and 50+ crate project it had slopped together lol.
You probably read too much into the downvotes, and people who downvoted you in the other thread might not even be the same people deciding on the future of C++.
I didn't downvote you (that's not something I do) but I didn't find your comment particularly relevant or on topic. This disapproval thing and working on memory safety don't seem related at all. Someone told you they had the impression that you were pushing for Rust without mentioning it. Your comment might have felt like you were backseat driving as well. More memory safety in C++ would be cool, but do you have actual ideas on how to achieve this? There are a lot of reasons people could have downvoted you.
You shouldn't worry too much about it though, just relax and forget about it. It's fine. Happens to the best of us.
It was about time. BTW, when this gem uses EGui instead of QT6 (nothing against QT, but once everything uses Rust, it will be no reason to keep it) it will shine.
Would love to see where all the folks from 5 years ago, breathlessly declaring it's futile to start a new browser project, are now and what they are doing with their lives.
Because it IS futile. Occam's Razor shaped hobby project, that what it is. Where is the market share? Nowhere, because existing browsers are good enough and no corporation pushes Ladybird.
Showing how far one can go with modern AI tooling does matter only a little when all you do is looking at Chromium/Firefox code and recompiling the logic of it in another language.
> Showing how far one can go with modern AI tooling does matter only a little when all you do is looking at Chromium/Firefox code and recompiling the logic of it in another language.
Is that what Ladybird does? Can you point to a real world example? From the blog post, it seems there are real world differences (look at the performance when resizing with heavy JS!).
You're also mixing arguments. The original arguments against ladybird (which I vividly remember as I loved following them back then) were that it would be impossible to arrive at something that works for all use cases. Ladybird would be able to render toy websites like hacker news, but nothing of worth.
> it's not 1980 anymore, browsers exist, and they have marketshare
Okay, sure. What was the market share for Chrome in 1999?
There are legitimate arguments against Ladybird. But pointing to the market share of a non-existent product, in a category on the verge of a paradigm shift from AI, is not one of them.
Ladybird has been really shaping up over the last month or so, I pull and build the latest bits every morning and use it for low stakes web browsing (HN, news sites, reddit, youtube, blogs, etc.) and it's getting very close to the point where I forget that it's a pre-alpha.
Very exciting to watch a completely new browser/browser engine come together. Reminds me of the early days of Mozilla and Gecko
Posted with Ladybird :)
I suspect in a year or so, an entire browser can be prompted by AI and then it won’t feel all that cool anymore. Unless AI hits a wall soon (unlikely).
I've actually used AI to create a WebKit port (not an original browser). The amazing thing is that it worked--but it still took months of issue triage. Even when the AI was working on its own, it had to spend enormous amounts of time just on compiling and testing. There are so many moving pieces, and each individual issue could mean the AI spinning for days.
Even if a future AI can drive web browser development completely autonomously, I suspect it will take a huge amount of real-world time because that's just the shape of the work no matter how "smart" you are.
Yep, I've been building a custom H2/H3 server for nearly 6 months now maxing my Claude 200 plan every week. Very pleased with the results but it's definitely not a quick or cheap task, it's constant decision fatigue and tons of rigorous testing on every piece in turn. Fuzzing, mutation testing, Miri, protocol conformance suites, H2load, actual networked runs, so on and so on.
I mean it worked, if scaling doesn’t stop and it can do this non stop at GPT Ultra speeds, this whole task will be done in weeks, maybe lesser. Doesn’t refute my point, ladybird is unfortunately going to be a pointless project soon. Though I guess if you have fun, who cares.
No, because the bottleneck isn't just model speed. The model has to literally compile the engine and run the massive test suites. Some of that could be sped up if you had a real build farm instead of my little computer, but there's a limit to how much you can parallelize things. And remember, I was porting WebKit, not remaking it, which would be orders of magnitude more work.
So you're not going to get to "build me an original web browser from scratch, make no mistakes" and get a new app the next morning or even next week, I don't think.
I'm doing a web browser from scratch, no web or js engine reused (only general-purpose libraries allowed), in my spare time/tokens on a Claude $100 sub.
After a month or so, many simple sites work, and now the bottleneck is the stupid amount of JS sites use for everything, and Cloudflare which doesn't like someone using a non-officially-approved browser.
Absolutely not a replacement for Chrome or Ladybird, but it can be done already.
AI is trained on human works. The preferences Ladybird expresses will influence those browsers.
Why is it that no conversation can be had about anything without AI being shoehorned into it, what did the parent comment even have to do with this
"[...] as far as we know, no other browser can do this today. Here’s a text animation while JavaScript blocks the main thread in 1.5 second bursts, with Ladybird in the top left:"
This is amazing!! I'm really surprised no other browser is able to do this. Is there an implicit "security" issue with it, or maybe standardization when it comes to keeping some values from DOM-Javascript in sync? Or simply an oversight in their part? In any case, congratulations to Ladybird's team. Computers are so much more capable of raw performance and computing, and it's nice that some developers out there do care heavily about it
afaik the decision to run animations on or off the main thread is a recurring point of debate not just in web browsers. On iOS for example SwiftUI runs them on the main thread while UIKit w/ CoreAnimation runs them off it. I don’t remember the specific reasons for these decisions but the point is, as with many things there are pros and cons to both approaches.
Opera's old engine (Presto) was able to do this too, before Opera switched to Blink. This is a perfect example of why browser diversity is important.
Ladybird is such an exciting project, really excited to try it out for day to day browsing when they get into Alpha.
It's very easy to build it yourself if you want to try it right now.
https://github.com/LadybirdBrowser/ladybird/blob/master/Docu...
I'd love a Flatpak... and maybe Gopher support in the browser ;)
Is it just me, or Google Maps is really broken in Ladybird?
I remember the first time I compiled it (IIRC in June), I was blown away, it was a bit slow, but it worked. (My standard browser performance test is Google Maps - zooming, panning, street view etc really shows the browser's capabilities.)
I just compiled the latest, and Google Maps is almost completely broken. You can kind of move around and zoom, but almost all the widgets are missing. The search bar is not missing, but it doesn't do anything.
I get that this project is still in development, but given the progress in the past few months, it's just strange that one of the most important pages on the internet is so disfunctional.
ps. Bing maps works much, much better. Is Google behind this?
Honest question for people. I'm a C++ dev and I'm seeing a lot of moves to rust. Everyone is using agents to move stuff to rust and I tried it myself. The amount of insane hardrive space usage (over 90GB of build artifacts that constantly need cleanup) is that normal? What am I doing wrong. Are people seeing advantages after a move to rust that wasn't there in C++ (assuming no new features). Like is your project better after moving to rust in obvious ways?
Once I spent a month debugging a UAF bug, in a messy C++ codebase.
With rust, those class of errors are very difficult to arrive at, unless you specifically make it so.
with rust, the "risk" gets moved from runtime to compiletime. I'll take a 1TB build bloat over code crashing in prod. any day.
I have the same problem and am regularly `cargo clean`ing hundreds of gigabytes of build artifacts. It is a real problem. I feel like I need minimum 4TB of SSD space just to comfortably work at the scale I'm working now in Rust. That's just for the working space of the target folders across worktrees.
I just think agents benefit a lot from having unavoidable feedback in the dev loop, and Rust affords that through its strictness. It's easier for agents to fall into the pit of success with Rust. I do think it's somewhat lacking in compile-time features, though.
It's expected. Cargo has very conservative cache cleanup (once per month IIRC).
There's incremental compilation cache that generates lots of churn (although Rust's compilation unit is a crate, "queries" in the compiler are cached individually, creating many cache entries per function per type).
The other half of the disk space is debug info, because abstractions that optimize down to zero runtime code still carry all of their source-level debug info.
These can be adjusted in build profiles if you're ok with coarser debug info or less fine-grained caching.
Cargo has no cache cleanup for `target` on stable.
It's normal. But you can somewhat mitigate by:
- Making sure you compile with as few profiles as possible.
- Disabling debug info (or setting it to "line-table-only") for the default profile.
It's normal, but they're working on it:
https://blog.rust-lang.org/inside-rust/2026/08/18/reducing-t...
Shared libraries (i.e. not building every damn dependency) would fix this problem real quick in many situations, but of course, Rust people don't want to commit to an ABI in order to keep their options open. I don't think it makes sense much longer - at this point "can't do shared libraries" seems more like cultural inertia.
Two or so ABI breaks with a few years between them also seem quite survivable.
90gb? i've ran cargo clean on a single project and reclaimed over 700gb before.
but i still love rust ofc
Jesus, that makes the JavaScript ecosystem look moderate by comparison!
The difference is that this accumulates over time. You'll only get this if you haven't run "clean" for some time (likely months).
it was days, but it was related to something the agent was doing with tests and targets, where it was basically compiling a billion variations of the almost 500,000loc, 1100 deps, and 50+ crate project it had slopped together lol.
Isn't the advantage memory safety?
When I mentioned, after C++26 failed its vote, that maybe they should use this extra time to focus more on memory safety, I got downvoted instead.
Surely that attitude isn't helping their situation.
You probably read too much into the downvotes, and people who downvoted you in the other thread might not even be the same people deciding on the future of C++.
I didn't downvote you (that's not something I do) but I didn't find your comment particularly relevant or on topic. This disapproval thing and working on memory safety don't seem related at all. Someone told you they had the impression that you were pushing for Rust without mentioning it. Your comment might have felt like you were backseat driving as well. More memory safety in C++ would be cool, but do you have actual ideas on how to achieve this? There are a lot of reasons people could have downvoted you.
You shouldn't worry too much about it though, just relax and forget about it. It's fine. Happens to the best of us.
[dead]
>Flatpak
It was about time. BTW, when this gem uses EGui instead of QT6 (nothing against QT, but once everything uses Rust, it will be no reason to keep it) it will shine.
Wow, they now beat Safari in tests passed. Catching up to Firefox too. Lots of momentum here.
Would love to see where all the folks from 5 years ago, breathlessly declaring it's futile to start a new browser project, are now and what they are doing with their lives.
Because it IS futile. Occam's Razor shaped hobby project, that what it is. Where is the market share? Nowhere, because existing browsers are good enough and no corporation pushes Ladybird.
Showing how far one can go with modern AI tooling does matter only a little when all you do is looking at Chromium/Firefox code and recompiling the logic of it in another language.
I'll stick with Firefox.
> Showing how far one can go with modern AI tooling does matter only a little when all you do is looking at Chromium/Firefox code and recompiling the logic of it in another language.
Is that what Ladybird does? Can you point to a real world example? From the blog post, it seems there are real world differences (look at the performance when resizing with heavy JS!).
You're also mixing arguments. The original arguments against ladybird (which I vividly remember as I loved following them back then) were that it would be impossible to arrive at something that works for all use cases. Ladybird would be able to render toy websites like hacker news, but nothing of worth.
It seems they were proven evidently wrong.
> Where is the market share?
Where was the market share for browsers in the 1980s?
But that's the thing: it's not 1980 anymore, browsers exist, and they have marketshare
> it's not 1980 anymore, browsers exist, and they have marketshare
Okay, sure. What was the market share for Chrome in 1999?
There are legitimate arguments against Ladybird. But pointing to the market share of a non-existent product, in a category on the verge of a paradigm shift from AI, is not one of them.
Waterfox switched their AD blocking engine to Rust. It's several times faster than UBo.
I will stop using the web entirely before I use anything made by Charlie Kirk-praising, DHH-defending, white supremacist-retweeting Andreas Kling.
[flagged]