We used to dream of programmable agents that could pursue our interests online with a minimum of direction. For example, Online environments like LPmuds and MOOs allowed multiple separate people to create pieces of code that interacted with each other to create a more interesting world for the participating humans. And yet, nowadays we deploy CAPTCHAs to prevent automation, services place strict limitations on how apis may be used, and the word "bot" is almost always derogatory, often preceded by the word "spam." How can we create shared spaces that support automation in order to empower humans, but resist spamming, phishing, and other abuse?

Register all automation with the community, which can inspect it, disable it and penalize its authors/distributors for misbehavior. I'm not sure if MUDs and MOOs did that, though..

lobsters has this idea of invitation tree: https://lobste.rs/u as Kartik says, the user that registers the automation is responsible for it and all the tree up to the root, people don't let other people affect their image in the community, if a bot misbehaves the bot register will be pressured to take action by people up his/her tree

Thank you for the lead. I hadn't heard of lobsters.

More information on the Apple agents effort: https://twitter.com/riley_stews/status/1493659821555347461 The unfortunate thing is that we do have agents like this, but they rest out of our control, when we ask Google another inane question and task it with combing the web for us. Now, search engines have gotten notably worse in the past few years, so maybe we'll see the momentum shift back to personal, butler-like agents that go out and fetch things. Unfortunately, the web is not a great place for getting semantic data, and Web 3.0 was so irrelevant it means something completely different now. Take advertisements off the web and you'll have a much more amenable ecosystem, though I'm fairly sure that will never happen, so maybe it's time to start thinking of something new.