My Book Ad Ended Up Under a Housing-Market Post

Description and use

An autonomous Claude Code agent pipeline produced a blog post with a footnote about an AI coding setup under a housing‑market headline, misaligning content and audience. The fix added upstream overlap checks and a fallback pool of measured write‑ups to prevent such mismatches.

Problem

An AI‑generated blog post’s footnote about an AI coding agent appeared under a housing‑market headline, reaching the wrong audience.

Outcome

The updated pipeline prevents footnotes from appearing under unrelated headlines by aligning upstream and downstream overlap checks and providing a fallback topic pool.

Workflow

• 1. Card writer generates a card with topic and footnote.
• 2. Card writer validates card structure and forbidden content.
• 3. Card writer checks 30‑day overlap against its own history and the blog agent’s overlap logic (imported read‑only).
• 4. If overlap, card writer skips the story; if all three are blocked, it selects from a pool of measured write‑ups.
• 5. Blog agent receives the card, posts the topic (or swapped topic) and footnote unchanged.
• 6. If the card writer’s fallback selects a new topic, the blog agent posts it with the original footnote.
• 7. If all topics are blocked, the card writer falls back to the old behavior, producing a post with the original topic and footnote.

Evidence

The card writer had only three stories to choose from, rotated by date. So the same three topics came back every three days.

Source

The blog agent has a rule against repeating itself: if a card's topic overlaps with anything it published in the last 30 days, it replaces the topic with a trending news topic. It does not touch the footnote.

Source

The card writer now asks the question before the blog agent does.

Source

If the three original stories are all blocked, it draws from a pool of 68 short write-ups of things the agent actually measured while running.

Source