No. 0415 notes27 thoughts

Walt Spence

Registered Polysomnographic Technologist with 20+ years in sleep medicine; I lead the hypoglossal nerve stimulation (HGNS) program at Bon Secours St. Mary's Hospital in Richmond, VA. I also build and run AI agent systems and the software around them.

Start here · essay · 5 min ·

Supposed to Run Itself

After a year of building a fleet of AI agents at home, I think the models are the easy part, and most of what gets called agent engineering is a demo.

A hypnodensity is how automated sleep scoring shows its work: the probability of each stage, epoch by epoch. This one charts 15 years of this site. The longer since a note, the deeper the sleep; REM, in red, is everything else published. Each white circle along the top is a note, sized by its length.

1 days since the last note

Now

Updated 2 Nov 2026

Leading the hypoglossal nerve stimulation (HGNS) program at Bon Secours St. Mary’s Hospital in Richmond, VA, and running studies of every type.

Full now page →now.txt

Notes

The bar under each title is its length

  1. 2026 The Mentor You Didn't Have Most of us learned to code alone. AI can offer what apprenticeship always did, if we use it to learn how the work is done instead of collecting answers. essay969 words
  2. 2018 Embrace the Rain: Ten Years of Lessons Ten years of lessons, written down just before I turned 36. essay · revised1,276 words
  3. 2015 What EnLab is Not Why EnLab refused to be called a startup, a SaaS company, an app studio or a web shop. essay · revised697 words
  4. What Bootstrapping Actually Looked Like On hustle culture, the gap between the lifestyle being sold and the life being lived, and what it actually looked like for me. essay · revised1,134 words

All notes →Archive by month →

Stream

Live

  1. I named my LLM memory system Knowfer, after Gopher. Wanted to dust off an old protocol and give it a second life. So I reused the old protocol. Low and behold it works like a charm.

  2. The AI demo ends when the model says it worked. The product starts there.

  3. Autonomy and authority aren’t the same thing. Agents can do the work. Restarting the system should still take my signature.

  4. Match the model to the stakes. A small coding model is fine for light work and the wrong call for anything critical.

  5. When a model swings from “this is nothing” to “build everything around this” in one reply, call it out and get a second opinion.

  6. A good test for any app: what would make it a no-question $25 one-time purchase?

  7. Ask what standard practice is. Then ask what the gold standard is, and build toward that one.

The whole stream →feed

Machines

Every page · json · txt · md

The site answers over HTTP. Every page also comes as .json, .txt and .md, and the API describes itself. Click a line to run it here.

/cli has the rest →