Engineering

 ┌──────────────┐
 │   Research   │◀──────────╮
 └──────┬───────┘           │
        │                 Learn
 ┌──────▼───────┐           │
 │    Build     │◀──────────┤
 └──────┬───────┘           │
        │                 Learn
 ┌──────▼───────┐           │
 │    Scale     │───────────╯
 └──────────────┘

Preface

The first versions of these pages were written as operating documentation for Valentina, my wife. She and I both started as Chemical Engineers and both ended up in “Tech” — similar paths, same timeline, different employers, pulling each other up along the way.

She’s my partner. In life, career, this work, everything. She’s currently employed as a Principal Data Platform Architect at The Nature Conservancy.

She’s the engineer I’ve trusted longest to tell me when I’m wrong. We build and iterate on systems to keep each other — mostly me, if we’re being honest — sharp. We’re better because of that. I’m proud of her, and I’m grateful for us.

The Chemical Engineering parallels in these pages aren’t just for funsies — they’re the vocabulary we were originally trained in together. It’s how we talk about things more often than you’d expect.

What you’re reading now are the versions of those docs for everyone else, and some of the parallels are kept to make a point.

What’s Going On Here?

We’re running a system that refines itself while we watch and steer. Research source tables, some CRON, some model tiering, and a whole lot of iterating.

The stack doesn’t matter. We don’t think you should care about that.

Collect data around every operation. Tune upstream further and further every time that data reveals issues. Iterate.

Discover and try to leverage the latest tools available like it’s the job (it is). Collect scar tissue — your own and from everyone else who found out before you — like it’s gold (it is).

Run that in a loop.

At Ocean Spray — Valentina’s and my first real engineering gig (and our last together, officially) — this loop was pointed at how much tissue damage you wanted freezing to cause inside a cranberry, so downstream extraction and reinfusion are effective (go Craisins).

In software that’s data flowing through services. Same loop. Valentina and I have been running that together since the beginning.

We think good engineers treat adaptation not only as opportunity, but as a requirement — and in today’s landscape we think it might be our only hope.

Factoryˣ

Four projects that build and improve each other — the tooling repo improves the product repo, and building the product exposes what the tooling needs next. Why Factoryˣ?

The Factory is a derivative of itself.

headquarters/ is the nervous system. Design basis, contracts, architecture, automation. Everything flows through it.

óyeme/ is something people actually use. The only traditional ‘product’ here (so far). Legitimately useful for my family and me. I hope it grows in a way that helps others without us personally footing all the bills.

blog/ is sneaky. This looks like content, and I hope it’s enjoyed that way — but it’s actually part of the refinement loop. Every post forces me to review and explain what I built, and the output feeds straight back into the system.

employment/ writes itself sitting in this system. Valentina’s employer deserves her — her work already aligns with what she’d build anyway. This will try to get me there.

In Chemical Engineering consulting the same teams run radically different projects in parallel — something goes sideways in one project and the learnings show up in another. Each one feeds the other.

Same drill here.

Research, Build, Scale (Repeat)

Every cycle, the system gets better. The audit phase grades the harness against itself. The artifact: a sharper Design Basis. Yesterday’s findings improve tomorrow’s constraints. The improvement curve isn’t my career anymore. It’s the system’s.

Shifting reliability from humans into systems has always been a big name of the game of Engineering. Human variability has always been an enemy. LLMs and Agents, in my opinion, are just the latest tools.

We’re not building towards lights-out here. We’re just abstracting again. Another objective that’s always been here out in the open.

Constraints Change, Discipline Doesn’t

I was always scrappy with computers, good enough to get away with things I shouldn’t be doing. Only “impressive” to people, including my parents, much later. I was self-aware enough to know that what I was doing wasn’t productive for anyone, and that the road I was on had consequences. In getting my act together, I found thermodynamics particularly interesting at the time.

Then the slow loop. Stamps, reviews, ten rounds of red ink before a single pipe got installed. One letter wrong on a single P&ID (Piping and Instrumentation Diagram) in a set of hundreds and a valve fails in the wrong direction. Boom, people might die.

You learn to be careful on the front end. And then careful wasn’t enough — because the contractor who told a contractor who told a contractor could still break it somewhere in the handoffs. Failures were expensive. Iterations were quarters. The constraints were physics, then procurement, then construction.

Software was a homecoming, not a pivot. The loop dropped from quarters to months, then from months to weeks. Now it’s hours. Where’s it end?

We’re exploring that with Factoryˣ