Blog

A browser that acts like a person, and fewer questions

Two changes that make agents easier to live with: they get further on the web, and they stop interrupting you for small things.

A browser that looks like yours

When an agent needs a website, it opens a cloud browser. That browser is now Google Chrome on a full desktop screen, with ordinary fonts and the agent's timezone, and it moves the mouse the way a person does. It also copes better with pages that used to trip it up, such as forms inside embedded frames and links that open a new tab.

Some sites still ask visitors to prove they are human. The agent makes one patient try. If the site does not let it through, it stops and asks for you instead of fighting the check. You see a card that says "Needs you" with the reason, open the live view of the browser, do the step yourself (a sign-in or a human check), and hand it back. The agent carries on from where you left it.

When a site wants proof you are human, the agent stops and asks for you. Once you have done the step, the card reads Handed back.

Questions only when they matter

Until now an agent could stop to ask before things like archiving an email or clicking "Unsubscribe", which is a lot of questions for a job you had already asked for. Your request is now the permission. An agent asks first only when the next step:

  • cannot be taken back, such as permanently deleting something,
  • leaves your account, such as sending an email or posting a message, or
  • costs money.

Reading, sorting, labelling and archiving go ahead without a card. When an agent does ask, the card says what it is about to do, so you can answer in one tap.

All posts