Comment by orourke
15 hours ago
I feel that getting anywhere into the neighborhood of “kind of working” for a project like this is noteworthy and a huge milestone. Maybe a better headline would be, however: Agents almost create a working browser.
15 hours ago
I feel that getting anywhere into the neighborhood of “kind of working” for a project like this is noteworthy and a huge milestone. Maybe a better headline would be, however: Agents almost create a working browser.
Yes, if Cursor claimed "We let autonomous agents run for weeks, and they produced millions of lines of code, and it kind of looks like a browser, and it kind of runs", then I wouldn't have written and published TFA.
But their claim wasn't so nuanced, it was "hundreds of agents can work on a single codebase autonomously for weeks and build an entire browser from scratch that works (kinda)". Considering the hand-holding that seems to have been required to get it to compile, this claim doesn't seem to hold up to scrutiny.
I've watched them today work in the new repo - https://github.com/wilson-anysphere/fastrender/tree/main , adding another 50k lines trying to optimize scroll/rendering performance (spoiler: not really)
At this point, its 1.5mlocs without the vendored crates (so basically excluding the js engine etc). If you compare that to Servo/Ladybird which are 300k locs each and actually happen to work, agents do love slinging slop.