Comment by JimDabell

22 days ago

Ten years from now, the agent layer will be the interface the majority of people use a computer through. Operating systems will become more agentic and absorb the application layer while platforms like Claude Cowork will try to become the omniapp. They’ll meet in the middle and it will be like Microsoft trying to fight Netscape’s view of the web as the omniapp all over again.

Apple will either capitalise on this by making their operating systems more agentic, or they will be reduced to nothing more than a hardware and media vendor.

I hope so. We're right on the cusp of having computers that actually are everything we ever wanted them to be, ever since scifi started describing devices that could do things for us. There's just a few pesky details left to iron out (who pays for it, insane power demand, opaque models, non-existent security, etc etc).

Things actually can "do what I mean, not what I say", now. Truly fascinating to see develop.

  • Ah yes. “Non-existent security” is only a pesky detail that will surely be ironed out.

    It’s not a critical flaw in the entirety of the LLM ecosystem that now the computers themselves can be tricked into doing things by asking in just the right way. Anything in the context might be a prompt injection attack, and there isn’t really any reliable solution to that but let’s hook everything up to it, and also give it the tools to do anything and everything.

    There is still a long way to go to securing these. Apple is, I think wisely, staying out of this arena until it’s solved, or at least less of a complete mess.

    • Yes, there are some flaws. The first airplanes also had some flaws, and crashed more often than they didn't. That doesn't change how incredible it is, while it's improving.

      Maybe, just maybe, this thing that was, until recently, just research papers, is not actually a finished product right now? Incredibly hot take, I know.

      1 reply →

I think you are right. In fact, if were a regular office worker today, a Claude subscription could possibly be the only piece of software you might need to open for days in a row to be productive. You can check messages, send messages, modify documents, create documents, do research, and so on. You could even have it check on news and forums for you (if they could be crawled that is).

  • I wouldn't call that productive, not even close if you are just sending AI replies, offloading all your tasks and doing nothing. This is what execs think we do, while every job has a lot of complexities that are hard to see from surface level. Belief that all work can be automatable is just a dream that execs have.

I really do not think much will change. In 10 years, computers will still look kinda like today and will be used like today. AI will be used to make drafts, but nothing more. For everything else, they are way too unreliable.

People hate to change habits, and many here overestimate the willingness and ability of, especially, older people to change how they use technology.

I don’t doubt the end goal.

My point is that it won’t be a ‘layer’ like it is now and the technology will be completely different from what we see as agents today.

In 10 years you probably wont own a PC if things go the way all the corporations want.

  • Possibly so in urban areas. Internet is already available everywhere. Sell dumb devices that can remotely log in to virtual devices. An LLM can connect to this virtual device and execute whatever action the user wants. Centralising compute resources this way means it's likely cheaper to offer huge compute to tons of users and so rather than buying a smartphone, you buy a monthly subscription to AI which can do everything your device does but you just need to speak or text to it. Sub includes cost of dumb device maintenance, securing the data you sent to the virtual device, etc.

    Personal Computing as a service. Let the computer think for you.