Yeah, totally agree. But I don't think they can keep it locked down any more, with the rise of computer use agents, the boundary between interfaces designed for humans and those designed for computers is becoming increasingly blurred.
Cloudflare and Datadog are already training models to detect agent-like mouse movements and DOM interactions. They'll lock this down easily, and we'll just end up in an ML vs ML arms race. It's still way too easy to fingerprint a script vs a real human just by looking at input latency patterns tbh
Last year I was talking with somebody about business plans and we were talking about using browser extensions for social news applications to get around paywalls, another part of the "lockdown economy".
I have an "image sorter" that ingests image galleries, it used to use a bookmarklet that cued a webcrawler but after a decade of having my time as a user wasted clicking on those Cloudflare thingies I finally had it get in the way of my webcrawler.
I was talking about options w/ Junie and it suggested, "why not write a browser extension?" and I'm like "is that hard?" and it says "no" and 20 minutes later I have a prototype browser extension.
So that plug-in you're talking about where an AI can see directly into your browser is another option, another is vibe-coding a custom browser extension. I think it was always pretty easy to write a browser extension but I had I done it myself my usual scholarly way I'd spend a few hours reading documentation before I started coding and never realized just how simple it is... Just the way people are discovering web crawlers and API integrations are easier than they thought.
There really isn't hours worth of material to put together a browser extension. You can read all the relevant docs in tops an hour, but you can follow the MDN tutorial on it to get the "inject some Javascript into pages with a url matching this pattern" level in less than 10 minutes.
Yeah, logging "Hello World" from a content script takes 10 minutes. But then you run into Manifest V3 service worker lifecycles, bypassing CSP on the host page, and syncing state across tabs, and suddenly your "quick one-hour project" turns into a completely burned weekend
