Editorial #2 — 18 Sep 2026 (awaiting Sam's POST ticks)

[HyperfocuSam]

Nothing here publishes. This branch stages Editorial #2 so that a POST tick becomes one merge, and a KILL becomes one git revert.

Branched from origin/main at 9347236c. Not merged into main, not deployed.

Main moved four commits while this was being built (PR 19, the bot-proof contact form), so origin/main was merged into this branch in 47c06e31. That direction was chosen deliberately: a rebase would have rewritten all eight post SHAs and invalidated the revert instructions below. Only public/feed.xml conflicted, on its <lastBuildDate> line, and all four generated artifacts were rebuilt from the merged sources rather than hand-resolved.

The nine commits

Each post is exactly one commit. All eight were dry-run reverted against the tip of this branch and every one applies cleanly.

#	commit	post	route
1	f82207f5	traffic-light-protocol-ai-safety (EN rewrite)	/blog/traffic-light-protocol-ai-safety/
2	d543e09d	ai-change-management-what-actually-works (EN rewrite + new H1)	/blog/ai-change-management-what-actually-works/
3	587d52c4	9-percent-copilot-licenses (EN rewrite)	/blog/9-percent-copilot-licenses/
4	6bba95d6	ai-maturity-trap-stuck-stage-one (EN rewrite + new H1)	/blog/ai-maturity-trap-stuck-stage-one/
5	46ddf8eb	microsoft-copilot-training-hong-kong-tc (new TC twin)	/blog/microsoft-copilot-training-hong-kong-tc/
6	5b88976b	hk-2026-budget-ai-training-tc (new TC twin)	/blog/hk-2026-budget-ai-training-tc/
7	c9f57c6f	how-to-design-ai-pioneer-program-tc (new TC twin)	/blog/how-to-design-ai-pioneer-program-tc/
8	16dbeff3	why-i-teach-copilot-not-chatgpt-tc (new TC twin)	/blog/why-i-teach-copilot-not-chatgpt-tc/
9	2e21c236	generated artifacts (sitemap, feed, llms-full, content.generated)	—
10	47c06e31	merge of origin/main (PR 19)	—

Dropping a KILLed post

git revert --no-edit <sha>
npm run prebuild     # commit 9 goes stale on its own; this rebuilds it

Commit 9 is the only shared commit. Reverting a post does not need it reverted, because prebuild rewrites all four generated files from the sources on every build.

What changed in index.js

Two titles and their excerpts, because both H1s changed and the slug did not:

• ai-change-management-what-actually-works — the field-report title dropped "70+ Organizations"; the excerpt followed it.
• ai-maturity-trap-stuck-stage-one — the old title asserted the "most companies stuck at Stage 1" distribution Sam deleted from the landers in 229bc1a0; the excerpt carried the same framing.

Eight further edits for the twins: four new language: 'zh-Hant' entries, and linkedPost added to the four English entries that had none. The pointer has to be on both sides or zhFunnelLinks.test.js fails the pair and the hreflang annotation is discarded.

Two calls worth a second look before merge

TC dates follow the English parent, not today. All 26 existing twin pairs in this repo share a date, no exceptions, so 2026-03-15 / 2026-03-14 / 2026-03-04 / 2025-12-30 were used rather than 2026-09-18. The consequence is that the four twins do not surface at the top of /blog. One sed flips them if that is the wrong call.

9-percent-copilot-licenses still carries Sam's open question. The 4% sustain-rate figure is on no record. The question is in the file as <!-- SAM: … -->, so markdown-to-jsx strips it and it cannot reach the page. It does ship inside the JS chunk as a string, because content.generated.js mirrors the whole file. Answer it and take the marker out before this merges.

Verification

npx jest            11 suites, 190 tests, all passing (after the merge)
npx eslint src      exit 0, no findings
npm run build       Compiled successfully

All eight routes are in public/sitemap.xml and build/sitemap.xml, each exactly once, and every twin pair carries bidirectional hreflang plus an x-default pointing at the English side. Post count went 100 to 104 (72 EN + 32 TC).