Loop HUX-L4 · Marketing — Huxberry.com · reports to Sherif + Nish

One publish-ready article a week, quality-gated.

Drafts one planned article a week from the content backlog, runs a 13-point SEO and brand checklist on its own work, and sends it for human review. Publishing only ever happens after sign-off.

Built & tested — pilotWave 3🕐 Weekly⚙️ Claude pipeline + verifier
Part 1 — the plain-english view

The content gap no GCC competitor has filled — filled weekly.

What it watches
  • The content backlog — 9 planned pieces, Gulf-climate sleep guide first (the biggest search gap in our market)
  • Its own drafts, against a 13-check verifier: SEO structure, keyword targets, brand tone, claim safety
When & how it speaks

One reviewed draft a week to Sherif/Nish. On approval, it's published to the shop blog — the publish step is gated, always.

Why it exists

Search demand for Gulf-climate sleep content (humidity, cooling, mattress choice for this region) is real and uncontested — none of our competitors has written it properly. A weekly rhythm compounds: a year of this is a content moat.

1,879
words in the first draft — the Gulf-climate pillar page, 13/13 checks passed
13
verifier checks — proven to actually fail on a deliberately broken draft, so a pass means something
9
articles in the researched backlog, prioritised by search-gap size

Weekly

One piece per week

✍️

Drafts

From the backlog, with keyword data where available

📋

13-point verifier

SEO + brand + claim-safety; failures rework automatically

💬

Review copy out

To Sherif/Nish for sign-off

Publish gated

To the shop blog only after approval

What it will never doNever publishes without sign-off. Unverifiable product claims are softened or cut in review — that already happened once (an exclusivity claim was softened until the brand can substantiate it).
Where it stands today · July 2026Built and tested end-to-end: pipeline + verifier (proven against a negative fixture — it catches bad drafts) + the first full draft at 13/13. In review, the judgment model softened one unverified brand claim and fixed FAQ phrasing, then re-verified. Awaiting Sherif's routing decision from the Wave-4 questions.
Part 2 — under the hood

How it's wired: systems, models, and the path a number takes.

For implementers and the technically curious. The full build sheet — verified queries, thresholds, and build notes — lives on the specs page.

StateContent backlog
9 researched pieces, prioritised
Stephux_l4_pipeline.py
Draft → verify → rework until 13/13
StepOpus review
Claim-safety and tone — it ships, so it's judged
DeliversReview copy
To Sherif/Nish
Human gateSign-off
Approve, edit, or reject
StepShopify publish
Gated write to the blog — after approval only
System we read Automated step State / memory Human decision
SystemRole in this loop
Shopify bloggated writeThe publish target — writes only on human approval.
Brand rulebookhux-brand-guidelinesTone and visual language checks inside the verifier.
SEO checklist13-point verifierStructure, keywords, internal links, claim safety — the quality gate that makes weekly cadence safe.
Hermesagent runtime on our serverThe scheduler that wakes the loop up. Each loop is a cron job under a Hermes profile; the planned bizops profile will host the business digests (IT loops run under vpsops).
Email renderer + gwsrender_email.pyAll digests pass through one shared renderer: Huxberry-branded HTML, tables for repeated rows, a coral 'needs your response' box when the loop has questions, and an arrow link on every record. Sent from the loops mailbox via the Google Workspace CLI.
Model / brainWhat it does here
Claude Opusthe judgment modelReviews alert wording, thresholds, and anything a human will read and act on. Post-Fable, Opus owns everything that ships.
GPT-5.5 via Codexthe bulk-work modelWrote and maintains the mechanical parts — SQL, diffing, digest assembly. Effectively free on our existing subscription, so routine cycles cost almost nothing.
Claude Fablethe designer (retired)Designed this loop's spec and encoded the judgment as thresholds and checklists before its retirement on 8 July 2026 — so cheaper models can run it without Fable-level reasoning.
State & memory

The backlog with per-piece status, plus keyword data from the search-watch loop as it comes online.

Delivery

Weekly review draft; published piece after sign-off.

Safety rails

Publish is the only write and it's human-gated; the verifier's claim-safety check plus Opus review keep unsubstantiated claims from ever reaching the blog.

Before it can run for real
  • Sherif's review routing confirmed (Wave-4 question)
  • Optional: L2 keyword data sharpens targeting as it arrives
Full build sheet →