Room
AI Work Portfolio
Built things, with what is true about each one.
These are systems I built, not proposals. Each page says plainly what the thing does, what has been measured about it, and where it stops. Where a thing is unfinished, private, or has never run against real load, its page says so on the page rather than in a footnote.
Several of these grew out of one method, the presuppositional LLM wiki, and the essays behind them explain why each exists and how it was made.
Who built them, and the licenses behind the clinical judgment in them, is on the resume.
-
Five Companies Run on Plain Text Files: Entity OS
Every corporate book has the same ingredients. I could not afford to have mine cleaned up, so I worked out how to keep them myself.
-
How I Review AI Drafts While Walking Instead of Reading Them at a Desk
A push that sends a draft to a reading app and a pull that folds my highlights and comments back into the same file. On one measured day: 74 comments and more than 16,000 steps.
-
No Source Gets Used Until Its Review Is Written: A Clinical Knowledge Base
Neutrality is a myth. Everyone has presuppositions. So every source that comes in gets read, written up and weighted before anything is taken from it.
-
One Article Through a Writing Line Built Like a Factory Floor: The Test Run
Research pulled when a paragraph needs it, with the inspections built into the line rather than bolted on after. One article has been through it, as a test that the line worked.
-
One Morning Page for Every Company I Run: Answer by Number
Future state, present state and what is waiting on me, for each business, on one page. The items are numbered and I answer by typing a number.