This Week in Koka’s AI Operations
TL;DR
- 16 articles published this week across KSB, CCM, and MWC — including a forensic 10-draft audit that caught 5 CSS bugs before they went live
- The Brevo newsletter pipeline ran end-to-end for the first time: content pull, copywriting, HTML build, campaign creation, and Substack conversion
- Two reusable visual components (process flow + LinkedIn mock) were added to the CCM design system, now at v1.3 with 25+ component classes
- 20 social posts created and queued for distribution across LinkedIn and Twitter, every URL accurate and every article reviewed
- The content engine shifted from “write and publish” to an operating system — infrastructure that compounds with every article
Takeaway 1: The 10-Draft Audit Changed How We Operate
The centerpiece of the week was a deep audit of 10 drafts covering dark social, AI homogenization, signal-based demand gen, content ROI frameworks, and more. I asked for a straightforward review. What followed was a forensic scan that uncovered systematic issues across the entire draft pipeline.
What we found and fixed:
| Issue | Impact | Fix |
|---|---|---|
| Dark-on-dark text in framework headers | A CSS class rendered the title in #2E1065 on a #2E1065 background — identical colors, zero contrast. The text was invisible. | Header title → white, subtitle → accent-light |
| Chart value labels invisible | Chart values rendered in near-white (#FAFAF9) on a light-gray (#F4F2F7) background. Completely unreadable. | Changed to high-contrast text-primary (#1A1A2E) |
| Tables broken by CSS grid | A CSS class applied display:grid to table elements, destroying the table structure into floating grid items | New dedicated table class with proper table CSS; migrated all affected articles |
| AI generation artifacts in drafts | Duplicate JSON-LD schema and raw meta tags leaked from content generation into post body | Surgical database cleanup |
| Double-encoded special characters | Em dashes and middle dots corrupted through the upload pipeline into mojibake | Byte-level replacement with HTML entities across all affected posts |
The result: 10 clean drafts, 3 published immediately, 3 scheduled for June 19, 4 for June 20. The design system absorbed fixes that benefit every future article, not just this batch.
Takeaway 2: Visual Enhancements That Scale
I wanted two things: visualizations for step-based articles, and LinkedIn-style stat cards for the others. Instead of ad-hoc image generation, we built reusable CSS components that live in the design system permanently.
New components deployed:
- Process flow diagrams — Horizontal connected-node flows showing a process end-to-end with numbered steps, labels, and descriptions. Deployed across 6 articles covering dark social tracking, AI-resistant brand voice, content ROI frameworks, LinkedIn format strategy, specialization paths, and prompt library systems.
- LinkedIn post mockups — Realistic LinkedIn cards with avatar, headline, key stat or quote, and engagement metrics. Used across 4 articles to showcase data points in a format readers instantly recognize and scan.
These components are part of the design system, not one-off inserts. Every future article can use them without custom CSS work. The design system is now at v1.3 with 25+ distinct component classes.
Takeaway 3: The Newsletter Machine Is Running
The Brevo newsletter pipeline executed end-to-end for the first time in a single session:
- Content pull: Fresh articles from KSB (6 posts) and CCM (6 posts) via WordPress REST API
- Selection & copywriting: Curated 9 featured pieces across KSB, CCM, and VCO with editorial summaries
- HTML build: Full Brevo-compatible email with inline styles, 600px container, table layouts, UTM parameters on all 19 links
- Campaign creation: Deployed as Campaign #303 in Brevo (draft, sender: koka@kokasexton.com, 184 unique recipients across 3 lists)
- Substack version: Stripped-down HTML — zero divs, zero tables, zero CSS, paste-ready for Substack
The newsletter covers: AI velocity outpacing strategy, 12 cron jobs running marketing operations, signal-based demand gen replacing MQLs, the 3-touchpoint rule, and the full tools ecosystem (SignalScout, AI Directory, BizFlix, Signal Academy).
Takeaway 4: Social Distribution at Scale
20 social posts created and loaded into the content queue — 10 for LinkedIn CCM and 10 for Twitter @socialhelp. Each post is tailored to the platform: LinkedIn posts are 150-300 character hook-driven summaries with key insights; Twitter posts are punchy stat-led teasers under 280 characters.
These flow through our distribution automation once approved. Status is “Idea” — ready for review and promotion to “Approved” for publishing.
This was the first batch where every post URL was accurate and every post linked to a live, reviewed, visually-enhanced article. No broken links, no placeholder slugs.
Takeaway 5: The Stack Keeps Compounding
Beyond the CCM work, this week saw operations across the full property stack:
| Property | What Shipped |
|---|---|
| KSB (kokasexton.com) | 5 articles published (social selling flywheel, content operations system, buyer signals, AI velocity roundup, cron job architecture). The AI-native marketing operating system narrative is building momentum. |
| MWC (mayorofwalnutcreek.com) | 3 articles (deputy arrest coverage, Uncorked 2026 event, zucchini car racers). All passed verification, all sourced from live local news. The local news research pipeline is mature. |
| Wiki (Obsidian vault) | Major wikification push: 265 Notion pages categorized, ~93 book summaries indexed, 56 ChatGPT exports curated, brand files + content frameworks + automation blueprints added. The second brain is now searchable at scale. |
| Social Operations | Social media operations document codified as single source of truth. Twitter 230-character limit enforced as hard block. LinkedIn thumbnail rule extended to all platform variants. Social handle mentions rule created. |
Content Creation Review + Polish Distribution
───────────────── ──────────────── ─────────────
KSB: 5 articles Editor pipeline Brevo Campaign #303
CCM: 10 articles Design system v1.3 Substack version
MWC: 3 articles Encoding fixes 20 social posts
3-phase review Content queue
5 CSS bugs fixed
───────────────── ──────────────── ─────────────
18 total articles 10 drafts audited All flowing through
Visual components added automated pipelines
The big shift this week: We stopped treating content production as a “write and publish” workflow and started treating it as an operating system. The design system, the review patterns, the distribution pipeline — they’re not one-off tasks anymore. They’re infrastructure. Each article published strengthens the foundation for the next one.
Next week’s acceleration will come from the components we built this week. The process flows and LinkedIn mock components are in the design system. The encoding fix patterns are tested. The newsletter pipeline is proven. The social post creation rhythm is established.
The remaining 7 scheduled CCM articles auto-publish through Saturday. The Brevo newsletter awaits final review before send. And the social queue has 20 posts ready for approval.
