If we do allow companies to block AI agents from accessing our own computers and data, then the users are to blame for falling again into another BigTech trap.
I think the current useful state of consumer LLMs is a temporary subsidy, and the incentives to add ads are too large. And that will change everything, even tools that should work for the user. I recently wrote a blog post on this: https://digitalseams.com/blog/the-ai-lifestyle-subsidy-is-go...
How long do you think it will take for this to meaningfully override Apple's share price?
Mathias Wandel (an ex Blackberry engineer) has a neat video where he explains exactly how that happened and the attitudes are strikingly similar to the ones today.
None of them actually care that they are Apple products or iWhatevers. Most couldn't tell you what the difference is between what they use and windows or android. They just know to go to the apple store to get things that work for them.
The entire Android ecosystem is built around people not paying for anything ever, where the iOS ecosystem, people pay for things. And honestly, it’s a way nicer experience.
Maybe users would rather keep their data safe than have it exfiltrated by a confused AI?
Can you actually even do that today? Not on iOS, I presume, definitely not on Android, at least not without hacking it six ways to Sunday with Tasker and Termux and API access to LLM providers.
(And no, firing Gemini and asking it to kindly maybe search your GMail doesn't count - because GMail is not the only e-mail provider, and GMail app is not the only e-mail client. If I want this to be possible with FastMail as provider and FairEmail as the app, it's hack o'clock again.)
Vendors all across the board really hate to give users useful features, because useful features tend to side-step their monetization efforts. And if history is any lesson, they'll find ways to constrain and shut down general-purpose use of LLMs. "Security" and "privacy" were the main fig leafs used in the past, so watch out for any new "improvements" there.
They need APIs for it to be efficient. For whatever reason they didn't choose to use accessibility tooling to automate agents, and we haven't written REST APIs for 20+ years - they're left hoping a newly designed protocol will fix it.
That surprises me too. It's arguably the only way forward that has a chance of surviving for more than a moment, because accessibility actually has a strong cultural and (occasionally) legal backing, so companies can't easily close that off.
LLM companies could easily have made a similar impact by leaning on accessibility tooling. Pushing companies to better support ARIA standards online would have made a huge impact for the better.
Heck, throw a little of that LLM money towards browser vendors to even better support ARIA - personally I'd love to see a proper API for directly accessing the accessibility tree.
If anything, they'd have the reverse impact, unfortunately. The thing is, the companies whose sites/apps/resources would be accessed by LLMs don't want this. That's the entire point of the article we're discussing.
All I'm saying is, accessibility is the only interoperability wedge they can't just close off, without a huge community backlash and in some cases because of compliance reasons.
Expedia makes sense, for example, and it'd be nice if they had more of a reason to improve accessibility on their site rather than building out and maintaining an API service specifically for MCP.