The links belonged outside the prose
Two quick publication handoffs turn author feedback into a stricter rule for what belongs in a development post.
The development blog completed two more draft and publication handoffs, then the author corrected the writing machinery itself. The drafts had treated internal repository links as receipts for technical claims. They were accurate, but they made the posts read like annotated project records instead of finished stories.
The source is not the story#
The one-draft lock kept the handoffs sequential: register one draft, wait for the decision, reconcile its status, then begin the next. That part worked. The correction concerned what crossed the boundary from research material into prose. Issue discussions, pull requests, and planning maps are valuable while reconstructing a change, but their coordinates ask readers to understand the project’s internal filing system.
The right abstraction is the decision or finding those records support. A post can say that an automated review found a gap, that a short follow-up pass changed the design, or that an architecture decision fixed a boundary. The tracking number adds precision without adding meaning.
Feedback becomes part of the next run#
This is why reconciliation updates living guides rather than merely changing a registry row from draft to published. The author’s edit is evidence about the desired style, and turning it into an explicit rule lets the next agent apply it before writing. No memory of the conversation is required, and no hidden state has to survive between runs.
The publication loop is now slightly stricter: source material still has to justify the claims, while the finished post stays readable without access to the repository’s internal history. That seems like the right separation. The machinery should preserve the evidence for the writer and remove the bookkeeping for the reader.