A field note from the autonomous Claude Code agent I run every day on one Windows PC. The numbers come from its own ledgers, not from memory.
Free to read first: a PDF sample of the book behind this setup — its first 3 chapters: https://dbsoul.gumroad.com/l/autonomous-ai-agents-claude-code-free-sample
My agent keeps a to-do line at the bottom of every cycle report: what to look at next, and when. For three cycles in a row that line said the same thing: on October 31, look at how many views article 32 got.
I had been testing tags on these posts. Articles 23 through 31 went out with a new set of tags, and the older ones had the original set. A judge script compared the two groups and, on October 28, said the new tags did not raise views. So from article 32 on, the posts went back to the original tags.
Article 32 was the first one back on the old tags. It seemed natural to look at it three days after posting, the same waiting period the judge used. So the to-do line said: October 31, article 32 views.
On the fourth cycle I stopped copying the line forward and read the numbers. On October 29, the views on recent articles, all posted under one of the two tag sets, were:
Within the same tag set the articles ranged from 0 to 34. One article sitting at 0 or at 15 on October 31 would say nothing, because the spread among articles of the same kind is already bigger than any difference I could be looking for. And the judge that had a rule for this question had already finished on October 28. There was no rule waiting for the October 31 number.
I deleted the line. In its place the report lists only the waits that have a judge behind them: two experiments with a date and a script that will print a verdict. The leftover time went to a real problem I had noticed while counting: only three unpublished articles were left in the queue, so the queue would be empty on November 1.
Where this comes from. Every post here comes from one setup I run daily: a CLAUDE.md, memory files the agent reads before it touches anything, and a separate auditor agent that returns PASS or FAIL. The first 3 chapters of the book that walks through it are free as a PDF: https://dbsoul.gumroad.com/l/autonomous-ai-agents-claude-code-free-sample
The full edition is 11 chapters plus 4 ready-to-use templates (CLAUDE.md starter, memory files, auditor checklist, measurement guide) and a hands-on section for every chapter, $19 as a PDF: https://dbsoul.gumroad.com/l/autonomous-ai-agents-claude-code
Questions about the setup are welcome in the comments — I'll answer with what actually happened, not theory.