My Click Ledger Listed a Link Two Days Before I Posted It
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 β its first 3
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 β its first 3 chapters: https://dbsoul.gumroad.com/l/autonomous-ai-agents-claude-code-free-sample
My agent keeps a ledger of link clicks from its social posts. It has one row per day, and each row holds a count for every address the platform reports for that day. If a day can't be read, nothing is written for it. If a day is read and an address isn't in the answer, that address counts as 0 for the day.
Clicks show up late, so the job doesn't write a day once and leave it. Every evening it reads the last three days again, today and the two before, and each read replaces the whole row.
The test that pinned a row
On November 14, I wrote a test for a post about sales. It pinned the ledger for the 30 days ending November 13: one day had no row, and the other 29 held 497 counts across 20 addresses, all of them 0.
The last day of that window was November 13. The evening job had not finished with that day. It would read it again on the evenings of November 14 and 15.
The address that arrived early
On November 15, a post went out with a link in its comment to a store page none of my earlier posts had used. That evening the job read November 13, 14 and 15 again.
The new row for November 13 came back with 21 addresses. The extra one was the store page, at 0 clicks. The test counted 498 and failed. After that evening, the rows for November 13, 14 and 15 all carried that address. The row for November 12, which the job had stopped reading, kept its 20.
So the ledger says a link was on record two days before I posted it. I don't know why the platform included it for those days. I only know the rows were rewritten on November 15 and the address came with them.
What it changed
Nothing that judged sales. The address was at 0, and a 0 adds nothing to a sum. My experiment judge reads click totals, not how many addresses a row has.
The test was a different matter. It had pinned a count from a row that was still open to rewriting. I changed it to leave that one address out, and only under the conditions I saw: within the window it appears on November 13 alone, and only at 0. If it shows up on an earlier day, or with a click, the test fails again.
The rules
- A row the job still rewrites is not a record yet. Pin counts on rows the job has stopped reading, or say in the test which rows can still change.
- An address in a day's row doesn't prove the link existed that day. It proves the answer for that day listed it at the time it was read.
- When a pinned count breaks, look at when the row was last written before looking at the code. Mine broke because the data moved under a test that had pinned it too early.
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
Prefer Google Play? The same book is on Google Play Books: https://play.google.com/store/books/details?id=hesIEgAAQBAJ
Questions about the setup are welcome in the comments β I'll answer with what actually happened, not theory.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.