I think a lot of people here are kind of concerned that there's only three browsers now, and overwhelmingly really only two: Google Chrome and Safari for iOS. Internet Explorer is just Chrome now and while I use Firefox it is still pretty tiny in usage stats. I do not consider things like Epiphany/Gnome Web as serious contenders, and even if I did that would still be only one more niche browser. Things without JS support like NetSurf, in my opinion, don't count at all.
It's a little scary when a single megacorp has so much power over something ostensibly open like the internet, but it also has historically taken an incredible amount of resources in order to make a browser making it hard for new players to break in. A modern web browser is arguably more complicated than an operating system...hell, it arguably is an operating system. It touches a ton of aspects of computer science, and requires lots of dedicated workers to keep up with web standards.
Because it has been such an intractable problem for so long, it's an extremely tempting target when the circumstances have changed. Anyone here can basically have a metaphorical intern working as many hours as they want for ~$20-$100/month. A problem that would be impossible for a single person five years ago suddenly seems "almost possible" when you can work at a higher level and have the pesky "code" details taken care of for you.
“A common complaint today from AI coding skeptics is that LLMs are fine for toy projects but can’t be used for anything large and serious.
I think within 3 years that will be comprehensively proven incorrect, to the point that it won’t even be controversial anymore.
I picked a web browser here because so much of the work building a browser involves writing code that has to conform to an enormous and daunting selection of both formal tests and informal websites-in-the-wild.
Coding agents are really good at tasks where you can define a concrete goal and then set them to work iterating in that direction.
A web browser is the most ambitious project I can think of that leans into those capabilities.”
“The browser and this project were co-developed and very symbiotic, only because the browser was a very useful objective for us to measure and iterate the progress of the harness. The goal was to iterate on and research the multi-agent harness—the browser was just the research example or objective.”
Because it's an extremely large and complex project that is also very clearly specified, to the point that the three word prompt "build a browser" encapsulates a huge amount of detail.
Similar to "build space invaders", another useful test prompt for seeing how well an LLM can do at a medium complexity task without having to give it a great deal of instruction.
It's a difficult, large problem - but one with very extensive tests available and that LLMs have a good understanding of. You can also look up the earlier automated ports of the JustHTML parser. A rendering engine is the next (admittedly large) step.
I say this as someone who also took on this task, about a week before I saw Cursor's attempt.
It would be very useful if AI could do it (though I kind of doubt something high-performance can be created this way). It's economically unfeasible for humans to develop a browser engine from scratch anymore. Microsoft and Opera gave up. Apple probably will too someday.
It's the most complex piece of software on your computer (it's basically an entire sandboxed operating system, at this point), the standards it has to adhere to are expanding every day, its performance optimization is critical yet adversarial (i.e. website owners have no incentive to make their sites efficient - the browser will be blamed for slowness, not them), and it costs the user nothing. No company can afford to maintain that, unless it's serving some broader strategy that's earning billions.
I've noticed a lot of desperation to validate the usefulness of AI. I guess partly from those who are trying to make money from it and partly from others who need the cognitive shortcut of offloading work and creativity to it.
Why did Alex Honnold climb Taipei 101 when he could just take the elevator?
I know the analogy is not perfect but it's the kind of project that wasn't feasible for a single dev before LLMs so now it just seems like a fun thing to try for some people.
And browsers are one of those products where if it was simple to alter everyone would like different setups and features. Like the Notes app of before.
Naturally they are all incomplete implementations because the AI agent is mostly reusing open source components and for the stuff it decides to write itself, doesn't have the training data to implement a full rendering engine in a one-shot.
Because the most used, high value target for exploitation application, is a perfect candidate for autocoded slop wrangled by incompetent developers who don't understand the code enough to write it from scratch.
I think a lot of people here are kind of concerned that there's only three browsers now, and overwhelmingly really only two: Google Chrome and Safari for iOS. Internet Explorer is just Chrome now and while I use Firefox it is still pretty tiny in usage stats. I do not consider things like Epiphany/Gnome Web as serious contenders, and even if I did that would still be only one more niche browser. Things without JS support like NetSurf, in my opinion, don't count at all.
It's a little scary when a single megacorp has so much power over something ostensibly open like the internet, but it also has historically taken an incredible amount of resources in order to make a browser making it hard for new players to break in. A modern web browser is arguably more complicated than an operating system...hell, it arguably is an operating system. It touches a ton of aspects of computer science, and requires lots of dedicated workers to keep up with web standards.
Because it has been such an intractable problem for so long, it's an extremely tempting target when the circumstances have changed. Anyone here can basically have a metaphorical intern working as many hours as they want for ~$20-$100/month. A problem that would be impossible for a single person five years ago suddenly seems "almost possible" when you can work at a higher level and have the pesky "code" details taken care of for you.
Blame Simon Willison ;)
“A common complaint today from AI coding skeptics is that LLMs are fine for toy projects but can’t be used for anything large and serious.
I think within 3 years that will be comprehensively proven incorrect, to the point that it won’t even be controversial anymore.
I picked a web browser here because so much of the work building a browser involves writing code that has to conform to an enormous and daunting selection of both formal tests and informal websites-in-the-wild.
Coding agents are really good at tasks where you can define a concrete goal and then set them to work iterating in that direction.
A web browser is the most ambitious project I can think of that leans into those capabilities.”
https://simonwillison.net/2026/Jan/8/llm-predictions-for-202...
“The browser and this project were co-developed and very symbiotic, only because the browser was a very useful objective for us to measure and iterate the progress of the harness. The goal was to iterate on and research the multi-agent harness—the browser was just the research example or objective.”
https://simonwillison.net/2026/Jan/23/fastrender/
Because it's an extremely large and complex project that is also very clearly specified, to the point that the three word prompt "build a browser" encapsulates a huge amount of detail.
Similar to "build space invaders", another useful test prompt for seeing how well an LLM can do at a medium complexity task without having to give it a great deal of instruction.
I called building a browser the "hello world" of complex parallel agent coding harnesses the other day: https://simonwillison.net/2026/Jan/23/fastrender/#a-single-e...
Because they can. It's fun.
AI makes it possible to do things you'd never have been able to do before, perhaps due to skill level or perhaps due to the time investment required.
It is a lot of fun to make software that you'd only been able to dream about making before AI.
I'd really like a pure RSS browser. Just nice articles in a standard format, no ads, no clickbait.
RSS readers are cool but the discovery is still tedious - looking for feeds isn't much fun.
I did just discover feedle though, which looks like halfway toward what I want! https://feedle.world/
I want feedle in a nice native app.
It's a difficult, large problem - but one with very extensive tests available and that LLMs have a good understanding of. You can also look up the earlier automated ports of the JustHTML parser. A rendering engine is the next (admittedly large) step.
I say this as someone who also took on this task, about a week before I saw Cursor's attempt.
It would be very useful if AI could do it (though I kind of doubt something high-performance can be created this way). It's economically unfeasible for humans to develop a browser engine from scratch anymore. Microsoft and Opera gave up. Apple probably will too someday.
It's the most complex piece of software on your computer (it's basically an entire sandboxed operating system, at this point), the standards it has to adhere to are expanding every day, its performance optimization is critical yet adversarial (i.e. website owners have no incentive to make their sites efficient - the browser will be blamed for slowness, not them), and it costs the user nothing. No company can afford to maintain that, unless it's serving some broader strategy that's earning billions.
Several years ago, I did write that every programmer should attempt to write a browser: https://austinhenley.com/blog/morechallengingprojects.html
:)
There's a lot of "if AI is so good then why can't it do X", and depending on who you ask "writing a browser" is somewhere on the frontier of X values.
It’s a demonstrably difficult problem that, aside from lady bird, has not been easy for independent devs to accomplish.
It is a stepping stone to writing an entire feature rich operating system thereafter, if only a virtual one.
I've noticed a lot of desperation to validate the usefulness of AI. I guess partly from those who are trying to make money from it and partly from others who need the cognitive shortcut of offloading work and creativity to it.
Why did Alex Honnold climb Taipei 101 when he could just take the elevator?
I know the analogy is not perfect but it's the kind of project that wasn't feasible for a single dev before LLMs so now it just seems like a fun thing to try for some people.
The same reason why all of a sudden people are rolling their own sandboxes using AI (without even looking at the code!):
Getting people to use it on the other hand...Because someone tried and it vaguely worked.
And browsers are one of those products where if it was simple to alter everyone would like different setups and features. Like the Notes app of before.
Naturally they are all incomplete implementations because the AI agent is mostly reusing open source components and for the stuff it decides to write itself, doesn't have the training data to implement a full rendering engine in a one-shot.
its just a meme
Because the most used, high value target for exploitation application, is a perfect candidate for autocoded slop wrangled by incompetent developers who don't understand the code enough to write it from scratch.
Yup. I think LLMs and AI in general can be useful, but I don't know what to call this wielding of it like a hammer in search of nails. A fad?