Echoes of the Machine: a framing for the next era

My first PC arrived Christmas morning, 1991, a 386 DX40 with 8 megabytes of RAM and a 120-megabyte hard drive that my parents had hidden in the family room until the reveal. I was in ninth grade. (There was a Commodore 64 before that, in sixth grade, but the 386 was the machine I actually moved into.) It was the kind of machine where you could feel the next upgrade coming. The kind where the era after this one always seemed to be one step away.

I have spent the thirty-some years since chasing the sound of those machines. The hum of platters. The smell of warm capacitors. The patience of waiting for something to compile. I have watched virtualization arrive and I have watched cloud arrive and I have watched the entire automation discipline reshape itself around APIs that did not exist when I started. Each time, the world got better at hiding the thing underneath. Each time, the people who knew what was running underneath got slightly rarer.

This blog is for what got left behind, and what's about to become the next echo.

What I think this era actually is

There are roughly four ways the working engineering community is talking about AI right now. Two of them are wrong. The third is interesting. The fourth is the framing this site is going to use.

The first wrong story is that AI is a productivity tool. It is, but if it were just a productivity tool, the entire conversation would be settled by now. Spell-check is a productivity tool. Stack Overflow is a productivity tool. Nobody writes essays about Stack Overflow.

The second wrong story is that AI is going to replace us. It isn't. It will replace specific tasks within specific roles in ways that will be uncomfortable for everyone involved, and it will create new roles that the AI isn't going to do. This has happened with every general-purpose technology and it is going to happen again. The framing of "they're coming for our jobs" is the framing of someone who hasn't lived through more than one of these.

The third story (interesting but incomplete) is that we are entering an era of personal AI. Where each of us has access to a model that knows our work, our preferences, our history, and operates as a kind of always-on assistant. This is true. It's also a small piece of a larger reframe, and treating it as the whole story misses what's structurally different about this era versus the last.

The fourth story (the one this site is going to use) is that AI is the moment the field has to start taking ownership seriously. Not ownership of the means of production. Ownership of the artifacts that encode what we know.

The foundation that lets a piece of software approximate how someone thinks is here. It's cheap, and it's getting cheaper. The questions that follow from that, who owns the artifact, who licenses it, who pays when it's used, who can copy it, what counts as derivative, are the actual interesting questions of this era.

What this site is

This is a personal blog about the technical foundation of the present moment. It carries forward the spine of a long-running cloud-automation site I wrote for years called dailyhypervisor.com, the full archive is here, dated as it was originally written, alongside the new work. Two corpora, one engineer, same shelf.

The cadence will be roughly two posts a week, more when something demands it. The pieces will be opinionated. They will sometimes be wrong, and when they are wrong I'll come back and write the second piece that says so.

Recurring threads

• AI as I'm actually using it, local-first, governance-aware, increasingly Apple-Silicon-shaped. Not as the vendors want me to talk about it.
• Cloud and automation, still the bones of everything, still worth understanding deeply. The category I built my career on doesn't go away because there's a new category on top of it.
• The ownership thread, a long arc on personal AI ownership: who owns the artifact when a model is shaped by a person's expertise, what a portable expertise object would actually look like, and where the supply-side and licensing infrastructure still has to be invented. This is going to be the spine; everything else hangs off it.
• Lived-in agentic operations, what working with these tools every day actually teaches you. The destructive cascades, the memory hygiene, the new metaphors that turn out to be useful. The kind of writing that the official documentation doesn't.
• Looking back, periodic pieces that connect the current moment to what came before. The 386 DX40 wasn't just a machine. It was a way of relating to the foundation. That mode of relating is getting rarer, and it's worth writing about why.

Why this matters

The people who lived through the last few computing-era transitions have something to say to the people living through this one. Not because the previous transitions tell us how this one will go (they won't) but because the texture of a transition is recognizable across them. The way the field temporarily abandons fundamentals. The way the loud voices win the early innings and the patient voices win the long game. The way the ground rules of ownership and licensing get rewritten in fights that the engineers building the systems don't see coming.

This is one of those transitions.

Echoes of the Machine is what got left behind in the previous ones, and what's about to become the next echo. Worth at least one place where someone writes it down honestly.

If you find something useful, let me know.

If you want to argue, even better.

, Sid