We will soon be publishing some insights from our approach. But for now, we enjoy the speed & accuracy of execution.
Most AI automation breaks when a button moves 2px left
@TryBrowzer, we switched from DOM selectors to the accessibility tree. 73% fewer failures in production
Turns out targeting what something *is* beats targeting where it sits on the page
PS: The video is NOT sped up
This guy is visibly pissed all the time
What I’ve really liked about @fdotinc is that, speaking with the team & other founders there has completely changed my perspective about content (in a good way).
I now think about content very differently than I did about 2 weeks ago. I’m visibly piss*d with meh content now
1
5
389
Before solving CUA, context aggregation needs to be solved.
Generic CUA is not very helpful for specific processes.
Learn the process & automate it following the process.
That’s what Browzer solves.
Computer use agents won’t work until there’s a context aggregation layer. Today’s agentic browsers and automation tools “assume” your processes or require you to “build workflows” and BOTH SUCK!
As cool as agentic browsers look, there’s no long term viability unless these browsers understand processes. NOT JUST WEBPAGES.
The difference between coding agents & CUA is that coding agents had readily available context. CUA doesn’t.
We need a frictionless context aggregation layer before we even think about reliable browser automation, and that’s what @TryBrowzer is building
1
5
789


