About Portfolio Cases Services Blog Contact 🎙 Talk to AI
EN DE RU
🎙 Talk to AI
Case study · Own product

Genesis Content OS

“An autonomous content system that reads its own audience metrics and opens pull requests to rewrite its own prompts — keeping only changes that improve real numbers.”

Created & built by Denis Shokhirev · Enterprise AI Architect, Freiburg
● Live Open-source · Apache-2.0 live.gerdennisai.com ↗
3,300+
Trend sources scanned
240
Articles in production
18
Machine-authored PRs
6 · 3
Channels · languages

Verified 2026-07-30 from production logs.

The problem

Manual AI-trend monitoring and multilingual publishing do not scale for a solo architect. Every hour spent watching feeds, drafting, translating and posting is an hour not spent building.

What I built

A closed content loop on n8n + Python + Supabase (pgvector) + Ghost + Astro, wired to GitHub auto-PRs. One topic record fans out to 6 channels — YouTube Shorts, Telegram, LinkedIn, Meta and Ghost in RU/EN/DE — with voice (ElevenLabs) and an avatar (HeyGen) generated from the same record.

n8nPythonSupabase (pgvector)GhostAstroGitHub auto-PRElevenLabsHeyGen
The part that makes it different

A self-optimizing module (“Module E”) reads the system's own audience metrics and opens pull requests that rewrite its own prompts. A measurement pass keeps a change only if real numbers (YouTube views) improve, and rolls back regressions automatically. To date it has opened 18 machine-authored PRs against its own prompt library.

Results (verified 2026-07-30)
  • Scans 3,300+ trend sources; publishes to 6 channels in 3 languages; 3 Ghost articles daily.
  • 240 articles live in production (in the sitemap).
  • Self-improving loop with automatic rollback on real metrics — not vanity dashboards.
  • Public live dashboard at live.gerdennisai.com. Open-source (Apache-2.0, patent grant).
Build one for your team

Autonomous systems, not demos.

If you have a real content or automation problem that needs a production system — let's talk.

Start a project → See all systems