Construction, Not Recovery
The series starts from a neuropsychological evaluation in July 2022. Processing speed came back in the 99th percentile. Recall memory came back in the 12th. Most AI writing assumes a working baseline and asks how to make it faster. I don't have that baseline, so these essays ask what a tool has to hold for the work to get done anyway.
Each essay ends with a claims list. Every claim is marked [SOURCED] or [GENERATED], and the known gaps are listed next to it.
The essays
Prosthetics, Not Amplifiers
Most AI productivity content assumes you're starting from a functioning baseline. I'm not, and that changes what the tools are for.
Claude Code at Ninety Days: What Held
Most of what I built in the first month, I rebuilt in month three, with the same tool.
Make vs n8n: Which One I'm Still Running
Sixty-five workflows marked active. The execution count came back zero on every one.
The MCPs That Earned Their Keep
Over a hundred MCPs connected. Three ran last month.
When the Stack Falls Over
Three silent failures in one month, and no alert for any of them.
The Claude Code Blueprint Nobody Posts
What the viral cheatsheets get right, what they skip, and my project structure after fifty skills and six months.
Fifty Skills and What They Actually Taught Me
The skills I planned were the least useful. The ones I captured from mistakes kept getting used.
New essays by email
About one a month. Nothing else gets sent.