The actual sequence of how I organized insurance for all four family members in one day with Claude—after learning from a registered letter from the district office that my father’s mandatory car insurance had been lapsed for a month. I gathered information scattered across six sources and turned it into a Google Sheets ledger, calendar schedule, and one-page card for my parents, while also showing the line where humans handle passwords, identity verification, payments, and consent all the way through.

We built a harness (rule files, blocking hooks, and completion gates) from 8 real-world incidents to prevent accidents where AI agents report “done” without verification, delete entire folders, or overwrite one another’s work across sessions—and we’re releasing it as a free starter kit. Alongside measured results from running the same instructions 16 times with and without the harness, we’ll show you in 10 minutes how to make AI’s “done” speak in verification numbers.

Based on the LLM Wiki architecture proposed by Andrej Karpathy, we are freely releasing a practical pipeline and MIT open-source starter kit that automatically compiles five scattered Markdown notes (8,615 characters) into nine wiki documents (23,659 characters) and a knowledge graph with 44 bidirectional [[wikilinks]] in just 1 minute and 42 seconds.