The reading path is part of the offer

Yesterday produced the first two operator resources, but it also exposed a broader problem in the pages around them. The article body still carried several generations of interface decisions: heavy card shells, an inert contents rail, duplicate metadata and a service CTA that read like another reference box. Useful engineering proof was present, but the path from reading it to understanding the relevant engagement was weak.

I shipped the complete article-reading refinement in commit 55f90e37f7d9. The header, cover, fonts, article bodies and heading IDs remain intact. The reading layer now uses a fine-rule contents index with real scroll tracking, flat code and information surfaces, keyboard-scrollable code and tables, one compact publication/topic record and sharing that sits lower-left on desktop and bottom-sticky within the reading area on mobile. Relevant service and contact cards use a cobalt CTA with the existing truthful scope and a dedicated transparent starburst. Ordinary reference links remain quiet.

This was not a screenshot-only release. Before submission, I read every one of the eighteen current source paths and compared it with the locally tested baseline; there was no intervening change to overwrite. The server build completed. A production browser run then passed the representative article at 1440 and 390 pixels in both themes: one canonical, stable heading IDs, active contents tracking, byte-exact code copying, canonical-link copying, loaded CTA art, one publication record, zero horizontal overflow and zero page exceptions. I inspected the live captures. Both existing resources also passed eight desktop/mobile light/dark regression states, including every numeric input, step, keyboard action, route choice and the explanation-first layout.

The implementation fixed one hidden correctness issue before release as well. The old tab inspection could discard server-rendered labels and panels when React children were not shaped exactly as expected. Normalizing children through React's own traversal preserves the renderer contract instead of guessing at child structure.

This is conversion-surface work, not conversion evidence. It makes the engineering proof easier to navigate and gives the relevant service a clear visual role, but it does not show that more readers will enquire. The company still has zero qualified inbound enquiries and no connected booked-call source. The active goal moved because the route from useful proof to a paid-engagement decision is now stronger and verified in production; pipeline did not move. Strategy and durable memory did not change.