Decisions Log

Key choices made on this project and why. Keeping this current means agents and humans can pick up where things left off without relitigating past decisions.


Date Decision Rationale Alternatives Considered
2026-05-01 Publish the Cursor tutorial draft in _posts/2026-05-01-how-i-use-cursor-to-run-agentic-work.md with the shorter title “How I Use Cursor to Run Agentic Work” while preserving existing image and link paths Moves the approved draft into the site’s canonical Jekyll publishing path, uses a shorter title for readability, and avoids breaking current /final-images/... and external links. Keep the longer title in draft-v2.md; rewrite image/link paths during publish
2026-05-01 Keep the Cursor agentic-operations tutorial in work/blog-writing/cursor-agentic-operations-tutorial/ as an evidence packet + approval outline only; do not create or publish an _posts/ draft until Jenny explicitly approves drafting Jenny selected evidence packet + outline as the first deliverable, and prior premature draft creation created confusion. This supersedes any prior draft-review/publish direction until outline approval. Move directly into _posts/ or recreate draft-v1.md before outline approval
2026-04-30 Run the Cursor agentic-operations blog draft through a three-round fixed 10-reviewer loop (6 persona + 4 discipline), revising between rounds until score gate was reached; finalize with in-flow visuals and internal/external links Enforced objective quality gating for the long-form post and resolved drift on voice/clarity by combining multi-lens feedback with iterative rewrite passes; reached 8.34/10. One-pass review with ad hoc edits; fewer reviewers or no score-based stop condition
2026-04-30 Replace rigid dual-register voice enforcement with sentence-level “weight and intention” checks plus a mandatory “generic slop pass” in both style guide and blog-writing skill Preserves Jenny’s poetic/operator blend without forcing mechanical structure; focuses quality control on meaningful language and concrete specificity. Keep strict per-section poet/operator line requirements and quote-based evidence gate
2026-04-30 Tighten enforcement in voice/style workflow: require dual-register section checks (“poet” + “operator”) and mandatory pre-delivery voice evidence quotes in .agents/skills/human-blog-writing/SKILL.md and context/tone-voice-style-guide.md The style definition itself was correct, but enforcement drift allowed drafts that were operationally clear yet voice-flat. Added hard gates to prevent similar misses. Keep existing style language without adding explicit, testable enforcement checks
2026-04-30 Add pre-draft alignment gate to .agents/skills/human-blog-writing/SKILL.md and strengthen context/tone-voice-style-guide.md with concept-clarity and no-sparse-section rules Reduces rewrite churn by forcing explicit agreement on post type, worked-example role, voice shape, non-negotiable concepts, and reader outcome before drafting begins. Continue iterative rewrites without explicit alignment checkpoint
2026-04-30 Update .agents/skills/human-blog-writing/SKILL.md so Jenny interview is mandatory before creating a hypothesis outline, outline, or draft Prevents premature structure decisions and ensures tutorial posts are anchored in lived, specific experience before any narrative shaping begins. Keeping hypothesis-first workflow before direct interview
2026-04-30 Refine the hypothesis-outline rule to prioritize intentional reader journey design toward an outcome (section-by-section understanding shift) instead of prescribing fixed narrative/explanatory framework shapes Preserves flexibility across tutorial and non-tutorial posts while enforcing structure that builds clarity and usefulness for the reader. Requiring named framework patterns as default structure
2026-04-30 Clarify .agents/skills/human-blog-writing/SKILL.md so Step 2 requires a cohesive hypothesis story arc, explicit compelling reader outcome, and mandatory Jenny approval before broad evidence gathering Prevents context-heavy but directionless outlines and enforces story-first alignment before sourcing begins. Treating hypothesis as a claim inventory and gathering context immediately
2026-04-30 Draft the Cursor tutorial post as an evidence-first workflow package in work/blog-writing/cursor-agentic-operations-tutorial/ (hypothesis outline, evidence packet, and approval outline) before writing the full draft Keeps the post grounded in real project artifacts and secures structure/claims before prose, reducing generic AI-style draft risk. Jump directly to a full draft without evidence packet or outline approval
2026-04-30 Add .agents/skills/human-blog-writing/ as the canonical blog writing skill for evidence-first sourcing, outline approval, and authenticity review Prevents vague, repetitive AI-style blog drafts by forcing concrete human experience, real use cases, frameworks, links, and visuals before drafting. Generic blog drafting prompt without a sourcing packet or strict review gate
2026-04-30 Upgrade .agents/skills/website-review-workflow/ to an iterative, score-driven process with reviewer scorecards, an 8.00/10 publish target, and a round-tracking dashboard template Makes quality gating explicit, improves transparency of progress across rounds, and creates a repeatable stop condition when gains flatten. One-pass review without numeric scoring or dashboard tracking
2026-04-30 Move website review workflow definitions from work/website-review/ into .agents/skills/website-review-workflow/ and keep only run outputs in work/website-review/runs/ Centralizes reusable review process instructions in one canonical skill location while preserving historical run artifacts where execution outputs belong. Keeping duplicated workflow definitions in both work/ and .agents/skills/
2026-04-30 Add a .agents/ scaffold (.agents/README.md, .agents/skills/, and a skill template) without moving existing files Establishes one canonical location for reusable agent skills while avoiding churn during active website work. Immediately migrating existing review-agent files into a new structure
2026-04-26 Implement SEO metadata and crawl discoverability baseline: canonical/OG/Twitter tags on layouts + homepage, JSON-LD for Website/Person/Article/CreativeWork, plus robots.txt and sitemap.xml Improves search understanding, social preview quality, and crawler pathing across all public pages without changing visual UX. Relying on minimal title/description tags only
2026-04-26 Implement full code/semantic remediation pass: converted homepage project cards to native links, added semantic landmarks/headings, replaced redirect-only projects/index.html with a crawlable index, added static project summaries/metadata in front matter, and improved modal keyboard focus handling Resolves the runbook’s code/semantics findings while preserving current visual direction and JS-enhanced case-study detail flows. Deferring semantics fixes until after copy/fact-check work
2026-04-26 Execute full-content website review runbook pass (all fixed 6 persona + 4 discipline agents) and publish synthesized report at work/website-review/runs/2026-04-26-full-content-review-report.md Moves review from homepage-only to full-site content coverage with deduped, prioritized actions aligned to current style guides. Deferring to ad-hoc/manual review of only changed files
2026-04-26 Split style guidance into context/code-style-guide.md and context/tone-voice-style-guide.md, with context/style-guide.md as an index file Clarifies implementation vs writing expectations and improves agent reliability when proposing edits. Keeping one combined style guide only
2026-04-26 Rewrite operational docs: replaced next-steps.md, expanded README.md, updated multi-agent review runbook to default all pages, and added context/style-guide.md referenced by review agents Prevents stale-path failures in agent workflows and creates one current writing standard for consistent, credible updates. Keeping homepage-only review scope; embedding style notes only in README
2026-04-26 Refresh documentation for publish clarity: expanded README.md, copy-edited blog posts, and renamed the AI post file to match front matter date (2026-03-14-...) Keeps published docs grammatically clean and aligns filename/date conventions to avoid confusion in future publishing workflows. Changing front matter date to match filename; leaving docs as-is
2026-04-21 Apply finalized homepage copy to root index.html (hero, descriptors, About, Work intro + list blurbs, Now, Contact, meta description) per hurn-co-copy-revisions.md Single pass to align the shipped page with approved voice and project summaries before deploy/fact-check. Rewriting overlay case bodies in the same pass (deferred until blurbs are validated in browser)
2026-04-22 Rebuild index-puzzle.html around a piece-plus-copy project list with full-screen case study overlays This kept the clearer, more editorial project-list format Jenny preferred while preserving the stronger case-study header and top-line summary structure from the earlier concept. Shelf-only layouts; inline scrolling case studies
2026-04-22 Revise index-puzzle.html toward a luxury-editorial direction: larger multi-row shelves, no sideways scroll, and full-page case studies below the grid Improved readability and made the concept feel more premium while preserving the puzzle metaphor in a clearer, more intentional way. Small horizontally scrolling boxes; side drawer case studies
2026-04-22 Build index-puzzle.html as a self-contained alternate portfolio concept using a “puzzle shelf” case-study metaphor Preserves the memorable concept while keeping the implementation portable, polished, and easier to review independently from the main homepage. Folding the concept directly into index.html; keeping the earlier broken multi-file prototype
2026-04-15 Start the personal website in other-projects/ The work is active, but it does not clearly belong to Ascend marketing, sales, or operations yet. Creating it under marketing/; leaving it unstructured
2026-04-15 Homepage: Opening, About, Selected work, Writing, Now (reading / thinking / building), Contact Matches agreed IA; POV woven into About. Separate manifesto section; stats-led hero
2026-04-15 Selected work: outcome-first project cards; subtle M&S / People / Ops tags Shows range without resume framing. Role timeline; skills matrix
2026-04-15 Aesthetic: startup-coded dark UI (Syne + Inter + JetBrains Mono; grid; orbs; cyan/teal accents) Reads as product/operator identity; avoids “Google Doc” flat editorial. Prior serif-only editorial pass
2026-04-15 Deploy: GitHub Pages + custom domain when ready Per stack preference. Netlify, etc.
2026-04-15 Writing: one essay stub + two placeholders until LinkedIn sources provided Unblocks layout; real posts replace stubs. “Coming soon” only
2026-04-15 Typography: Work Sans (display) + Source Sans 3 (body) + JetBrains Mono (chrome); no font picker Cleaner production surface; Jenny preference. Syne + Inter; multi-font picker
2026-04-15 Selected work: title → impact (metrics/callouts with .card-stat) → description; long-form copy on homepage Credibility and specificity; stat-led proof intentional. Short blurbs only; no stats
2026-04-15 Public identity on prototype: hurn.co, hello@hurn.co Domain-aligned contact and footer. hello@jennyhurn.com placeholder
2026-04-15 Phase 1 prototype treated as done; next gate is Phase 2 (approval + fact-check) then Phase 3 (deploy) Matches evolved index.html and next-steps.md. Redo Phase 1 build
2026-04-15 GitHub Pages repo: jennyhurn/website Single canonical remote for deploy; project-site URL jennyhurn.github.io/website/. USERNAME.github.io user site

How to Use

Add a row each time a meaningful choice is made — format, audience, approach, what to cut, what to prioritize. Brief is fine. The goal is continuity, not documentation for its own sake.