Quick thoughts on personal agents
The rise of personal agents has been interesting for me. I spent the better part of this year building one for myself on Nanoclaw, so it's really incredible to see the broader progress.
It has also got me thinking about what's happening. Here are some of those thoughts:
-
They are expensive to run in their current state. Every user has a VM in the cloud - that's difficult to scale to a billion. But it's a solvable problem and I think the bet everyone is taking is that they'll solve it before it becomes a bottleneck.
-
Apart from the hype on Twitter, most people still don't know about or care for them. This means two things: a) the opportunity is still very large. b) distribution is key as it is for all products.
-
Trust grows with value. My personal email was a no-go for AI until Gemini Spark launched. Instinct's CEO also mentioned recently that 40% of their users give the agents a credit card to buy stuff within three weeks. As AI becomes more useful, the definition of "off-limits" for each user will continue to change.
-
While big labs will continue to focus on general purpose agents, there's huge opportunity for specialized vertical ones. And they'll even be able to talk to the general purpose ones to execute on tasks they do best. Expect the rise of "Instinct for X" startups.
Appeared first on my LinkedIn on Oct 1, 2026.