Eight posts, sixteen handoffs

The blog backfill has moved from a repeatable rhythm to sustained use: eight posts completed the draft-review-publish loop in one evening. They cover design attempts, infrastructure changes, launch work, reversals, notifications, and engine seams, but this part of the git history is about the publishing machinery itself. The earlier three-post run established the handoff; this run shows that the same constraint still holds when the queue is long.

Keep the lock, shorten the wait#

Every post still required two separate changes in the game repository. Opening a blog PR appended a draft row, merging it allowed a reconciliation pass to change that row to published, and only then could the next range begin. Sixteen bookkeeping PRs for eight posts is deliberately repetitive. Combining them would be faster on paper, but it would also remove the visible point where I approve one subject before an agent starts writing the next.

That result is useful, but the number needs careful interpretation. Eight clean merges do not prove that review can be skipped; they show that the guides and the one-draft lock are producing drafts I can approve quickly across very different subjects. The evidence is the repeated handoff, not the absence of corrections. A post about a design reversal and one about deployment infrastructure both passed through the same small protocol without being forced into the same narrative shape.

Repetition is part of the design#

The registry commits are intentionally boring. Each one leaves an auditable answer to three questions: which git range the post covers, whether the editorial decision is still pending, and where that decision happened. This is the kind of duplication that earns its place because it makes concurrency visible. An agent does not need hidden state or conversational memory to know whether it may proceed; it reads the tail row.

There is still more history to cover, and speed is not the goal by itself. The useful target is a short cycle that preserves the human decision, keeps ranges contiguous, and turns corrections into rules only when corrections actually occur. For now, sixteen handoffs without a broken lock is a better measure than eight posts in an evening.