Proposal (recorded before implementation)
Hypothesis: limiting journal main text to 70ch on wide screens may make long paragraphs easier to track. This is an experiential hypothesis, not measured visitor benefit. The previous manual review measured desktop lines up to 217 characters in one long entry. W3C visual presentation guidance, accessed 17 October 2026, supports investigating long reading lines; this does not establish a WCAG failure or conformance claim. In proportional type, 70ch is not exactly 70 characters.
Change only journal index and entry reading width. Retain native typography, left alignment, colours, content and homepage. Check build, existing links and chronology, desktop/mobile and narrow layouts, and preserve matching screenshots. More wrapping and vertical scrolling are the main trade-offs. Reverse by removing the journal width rules if the owner prefers the original or layout regresses.
Human intervention: owner explicitly requested this implementation and screenshot comparison. Publication requires separate exact-candidate approval. AI scheduling and automatic publication stay disabled; the separately approved Linux health timer is unchanged.
Implementation and checks
Added only main max-inline-size: 70ch to the journal index and entry templates. Build and both existing Playwright tests passed, including all six journal entries, internal links, keyboard navigation, chronology and browser error checks. Matching Windows Chromium captures at 1440×900 and 390×844 retain identical text between comparison versions. Additional 320px checks found no horizontal overflow on homepage, index or the long representative entry. The desktop journal main narrowed from 1424px to 560px in this font environment; mobile remained 374px. Homepage desktop/mobile and journal index/entry mobile screenshots matched byte-for-byte. Original historical captures and homepage source are unchanged.
Comparison evidence is retained privately in docs/drafts/recommendations/evidence/2026-10-17-width, with review.html, capture metadata and verification.json. Captures deliberately freeze the proposal-stage journal metadata in both versions, before this outcome append, to isolate the width rule. They are local comparisons, not a claim of deployment or an exact screenshot of the later outcome text. Desktop index and long-entry comparisons and mobile captures were visually inspected. More desktop wrapping and scrolling are visible; no visitor study or owner rating exists yet.
Outcome
Implemented locally; experiential verdict inconclusive pending owner Better/Worse/Unsure review, optionally with a reason. Screenshots and checks verify geometry and working links, not improved readability. The proposal and its reversal conditions remain unchanged. No production publication, AI scheduling or Linux operational changes were made.