Agents getting slower because of browser checks

Ideas Feedback
, ,
1

I use Grok 4.6 Fast High for most of my work. But I’ve noticed that it has become a lot slower on frontend tasks.

One reason seems to be that it spends a huge amount of time driving the browser, taking screenshots, and reviewing its own work.

I’m not sure if there’s a better way to do this. But the experience doesn’t feel good.

Before: about 1 minute to finish the work. Fast and satisfying. Even if there were small issues, I could iterate quickly in that loop.

Now: 1 minute of actual work, plus about 5 minutes of browser ops, review, and checking. A lot of waiting.

At that point, the experience is no different from Claude or Codex for me — meaning, it’s just slow.

I’m not saying verification is bad. I just miss the old speed. For frontend work, I’d rather get a fast first pass and iterate myself than wait several extra minutes for screenshot reviews.

8

Hey @Z_W, thanks for the report, this is useful feedback!

The agent will respect an explicit instruction here. Add a rule like “Do not verify changes in the browser or take screenshots unless I explicitly ask. Give me a fast first pass, and I will review it myself.”, and you should get back close to the old loop. For one-off tasks, saying “skip browser verification” in the prompt works too.

You can also turn off Browser Automation under Cursor Settings > Browser & Network to disable the verification altogether.

Hope that helps!