Back in 1977, the head of a big computer company said nobody would ever want a computer in their home. Within just ten years, he couldn't have been more wrong. Personal computers were everywhere.
What happened? We took computers, which used to be these massive things only big institutions could use, and made them personal. We turned them into tools anyone could use. Brain bicycles. This changed everything about how we thought about computers – they went from specialized tools to everyday ones.
Something similar is happening right now.
Just like we made computers work for humans in the 70s, we're now making computers work for AI. AI is starting to use computers like we do. But here's the interesting difference: when we built computers for humans, we could count on human common sense. We can't do that with AI.
A computer. For computers.
AI can now use computers like we do – clicking around, reading documents, doing complex tasks. And just like we built robots with arms and legs because that's what works in a world built for humans, AI needs to use the same kind of computer interfaces we use – windows, browsers, all that stuff. Not just raw code.
This brings us to the big question: how do you let AI agents be useful while making sure they don't do anything they shouldn't?
Some companies will try to solve this by just limiting what AI can do. That's the wrong approach. The winners will be the ones who build the right foundation.
Prelude to Foundation
Just like Microsoft won by building the operating system that made personal computers actually useful, AI agents need a secure computing environment that provides:
- Complete control over what the AI can and can't do
- A comprehensive record of everything it does (like a security camera for AI actions)
- Fine-grained control over what resources it can use
- An easy way to deploy and manage lots of ephemeral environments for AIs
Without this foundation, only huge tech companies with massive security teams will be able to use AI agents safely. With it, any company can. If we don't build this foundation:
- Security problems are guaranteed
- AI could use up all your resources
- You won't know what the AI did
- Setting up agents will stay complicated and dangerous
Starting Now
The technology exists now. From this point on, every company using AI agents will either build on a secure foundation, or they won't deploy agents at all and will be left behind. We're building that foundation.