Hi, I'm Matthew, a Go-to-Market Engineer with a passion for AI infrastructure. I spend my days at aetherfs.io, where we're building what we think is the missing layer underneath autonomous agents. This blog is not about that, or not directly.

This is the place I started writing for myself. After one too many conversations with co-workers at aetherfs.io, where someone said something genuinely surprising about agent infrastructure, I realized none of it was making it past our internal Slack. We argue about why the per-VM model is quietly breaking down. About what a session ID wants to be. About the strange fact that every serious agent platform has built its own filesystem internally, badly, in parallel. The conversations were better than anything I had been reading, and most of it was getting lost in scroll-back.

The series you'll find here is my attempt to put those conversations on paper. The voice is analytical, not AetherFS sales copy. I name competitors. I take positions that would be commercially convenient for me to soften, and I don't soften them. Two of the pieces are outright thought experiments arguing against the current consensus, and I run them knowing some readers will wave them off. I am writing the kind of blog I wish existed when I'm trying to make sense of a new infrastructure category, before I have any idea which vendors to trust.

If you're a CTO, principal engineer, or platform lead trying to figure out how agent state actually works in production, I think you'll find something useful here. If you end up at AetherFS as a result, that's good for me, and I won't pretend otherwise. If you end up convinced the category is real and someone else is the right partner, that's fine too. I would rather you read honest thinking and reach your own conclusions than read marketing and resent me for it later.