Deep Roots: Before the Cluster, There Was Dial-Up

Most origin stories start with a job title. Mine started with a question at a family friend's house: "D'you have a 32-bit CPU?"

Late-night diners: the original co-working spaces for broke tech enthusiasts.

The Context: Mid-90s Arizona

It was 1995, maybe earlier. I was in junior high, obsessed with computers. I had just convinced my parents to get a dial-up connection. Not just any connection—GetNet, a local Phoenix ISP. I still remember the logo: a little guy made out of two serial cables connecting as the head and arms.

We started with SLIP, then upgraded to the newer PPP protocol. Shell access. Usenet. A 14.4k modem screaming into the void.

I was also on the list for the Windows 95 Beta. This felt like being admitted to a secret society. A new operating system! Long filenames! Plug and Play!

I was at the home of a family friend—a small publishing house that produced educational typing software. One of their contract developers was there, a PhD candidate from ASU named Alan.

We got to talking about the beta. I was excited. He was… curious.

“D’you have a 32-bit CPU?”

I did. A 486.

“Will you join my cluster?”

I didn’t know what a cluster was, exactly. But I said yes.

The Lab: 10Base-T and Token Ring

That “yes” led to afternoons and weekends at the ASU computer lab. My job, such as it was, was informal: I could crawl under tables (useful for cable runs), I could follow instructions, and I was happy to be there.

The lab ran 10Base-T Token Ring. Windows NT workstations. Old hardware by modern standards, but it was real infrastructure. I learned the difference between a hub and a switch. I learned why you don’t unplug a Token Ring cable mid-session.

We installed operating systems via floppy chain—one machine finishes with disk 1, you’d walk it to the next machine, start the install, then circle back. It was a physical, human process. You felt every byte.

And then there were the breaks. Alan once pulled up a Unix workstation, fired up the Mosaic browser, and showed me a grainy video of a small town trying to dispose of a beached whale with dynamite. The famous exploding whale. It was absurd. It was glorious. It was the first time I understood that the internet wasn’t just for research papers—it was a place where anything could exist.

Alan was working on his PhD thesis: a Single Address Space Operating System called “Sombrero.” The idea was radical—a distributed OS where every object in the system could be addressed with a single, globally unique identifier, no matter which node it lived on. It required deep access to the NT kernel. There was a workstation in a secure pod—a secure room inside a secure building—that only he and his advisor could access. It held the Windows NT source code.

I wasn’t allowed in that room. But I learned about it. I learned that the software we ran every day was built by people, on purpose, and that the decisions they made had consequences.

The University of Denny’s

The real education happened after hours. After the lab closed, we’d pile into cars and head to Denny’s—the one at Mill and Guadalupe, or the one on Baseline. These weren’t “meetings.” They were late-night rambles over coffee and pie, dissecting operating system theory, network architecture, and why certain abstractions were elegant and others were traps.

I didn’t always follow the conversation. I was fifteen. But I absorbed it.

The lesson that stuck: Stability is architecture. A system that works under pressure isn’t lucky—it was designed that way. A system that fails was designed that way, too, even if the designer didn’t realize it.

Why This Matters Now

Twenty-plus years later, I still think about those late nights. The problems have changed—cloud, containers, zero trust—but the principle hasn’t.

Infrastructure should behave. If it doesn’t, the fix isn’t more heroics. The fix is better architecture.

I got my start crawling under tables and listening to people smarter than me argue about kernel design. The tools are different now. The mindset is the same.