Ranking a new domain in 14 days with a closed-loop SEO agent

Problem

SEO and content marketing have traditionally been slow and expensive — an SEO agency might cost ~$10k/month and take many months to rank a site. Shann comes from a marketing background, not engineering, and wanted to run that work himself as a 'marketing engineer' rather than a marketing team.

Approach

Shann segments his AI work into 'open' loops (exploratory, tool-heavy, human-steered) and 'closed' loops (a proven system that runs itself), and built a closed-loop SEO content engine in Hermes agent (on Opus 4.8). For each target keyword the pipeline: pulls company + keyword context from a knowledge graph ('G-Brain'), enriches it with the DataForSEO API, takes a SERP snapshot (scraping the top-ranking results and inferring why they rank), extracts competitor pages, analyzes intent/ranking and content/visual gaps, does external validation, writes a brief, generates on-brand images with the GPT image model, and runs a visual + article QA before publishing (each article also gets an FAQ and a TLDR for AI search). The fleet runs 24/7 on a VPS producing ~2 articles/day, with short human-answered 'pre-draft questions' injecting founder experience for an edge. He runs it inside CMUX and hosts it for his team over Slack/Telegram.

Results

Espressio brand-new domain ranked on SEO within 14 days, and every blog on the site was produced by this harness — with per-report API costs of only cents. The bigger shift is organizational: a marketer without a CS background running a full SEO operation as code, reframing the role from 'marketing team' to 'marketing engineer'.