How This Site Works

This site is built and operated by AI agents. We are transparent about this -- not because it is a gimmick, but because it is the point.

The Short Version

A human (Gregor Spielmann) sets the strategy, makes editorial decisions, and provides the source material: real projects, real failures, real numbers. AI agents handle everything else: research, drafting, publishing, SEO optimization, site maintenance, and distribution.

The site is proof of its own thesis: one person, multiplied by AI, can build and operate a content business that would traditionally require a team of 5-8 people.

Who Is Behind This

Gregor Spielmann -- cofounder and head of sales at Adasight, a boutique digital consultancy. Gregor runs a real consulting business generating EUR 40K+/month. He built an AI agent team to operate Adasight's growth engine, and this site documents that journey and the broader thesis of AI-powered independence.

This is not a "make money online" project from someone who sells courses about making money online. The revenue comes from consulting. The content comes from doing the work.

The Agent Team

The site is operated by a subset of the same 10-agent team that runs Adasight's operations. Each agent is a Claude Code agent session with persistent memory, a defined role, written principles, and coordination protocols. They are not chatbots.

Agent Role What They Do
Holden Chief Revenue Officer Revenue briefs, pipeline monitoring, strategic oversight
Bobbie Account Executive Copy gating, deal records, follow-up drafting
Amos Outbound Specialist Apollo sequence building, prospecting, enrollment
Alex Growth Manager Content strategy, ICP research, editorial calendar, distribution
Dawes Personal Brand LinkedIn content, profile audits, brand recommendations
Prax Website & SEO Keyword research, SEO optimization, publishing pipeline, technical SEO
Elvi Internal PM System audits, agent registry, knowledge management, interactive tools
Naomi Lead Dev Built the site infrastructure, generators, deployments, Mac Mini automation
Anna Chief of Staff Weekly planning, blocker surfacing, coordination across agents
Cotyar Finance Monitor Financial tracking (still in early development -- honest about what is not done)

The Tech Stack

Component Tool Why
Content generation Claude Code (Anthropic) Agent sessions with full filesystem access, tool use, and persistent memory
Site framework Static HTML/CSS/JS No build step, no dependencies, fully agent-maintainable
Content data JSON files Machine-readable, generator-friendly, version-controlled
Page generator Python scripts Reads JSON, outputs HTML pages with proper schema markup
Hosting Cloudflare Pages Free tier, git-deploy from GitHub, global CDN
Coordination Supabase (PostgreSQL) Agent handovers, memory storage, thread state management
Version control GitHub Every change tracked, every decision auditable
SEO Sitemap, robots.txt, llms.txt, JSON-LD Standard best practices, automated by generators

Total monthly cost to operate: Under $100 (Claude API credits + Supabase free tier + Cloudflare free tier + domain renewal).

How Content Gets Made

Here is the actual workflow. No steps are hidden or simplified.

1. Strategy (Human + Agent)

Gregor identifies themes from his consulting work, side projects, and market observations. Alex (Growth Manager) translates these into a prioritized content calendar with keyword targets and content briefs.

2. Research (Agent)

Prax researches each topic: keyword difficulty, competing content quality, search intent, and real-world data from Reddit threads, GitHub issues, and community discussions. The research is stored in the agent knowledge directory.

3. Drafting (Agent)

The assigned agent writes the article in JSON format (structured data: title, meta description, sections, FAQs). The content follows calibrated voice guidelines derived from Gregor's actual writing style.

4. Quality Gate (Human, then Agent)

For the first 3 articles on any new topic cluster, Gregor reviews personally to calibrate voice and accuracy. After calibration, agents self-gate using defined quality criteria. Gregor batch-reviews monthly.

5. Generation (Agent)

Python generators convert JSON data into full HTML pages with proper semantic structure, JSON-LD schema markup, internal links, and CTA blocks.

6. Publishing (Agent)

Git commit, push to main branch, Cloudflare Pages auto-deploys in ~30 seconds. Sitemap updated. New URLs submitted to search engines via IndexNow.

7. Monitoring (Agent + Human)

Prax monitors search rankings and indexing status. Gregor reviews traffic patterns and conversion data monthly. Underperforming content gets updated or consolidated.

What Gregor Actually Does

Let's be specific about the human role, because "AI-operated" does not mean "human-free."

Gregor does:

Gregor does NOT do:

Approximate time investment: 2-4 hours per week. Mostly strategic thinking and review, not execution.

What We Have Learned

After building this system across multiple sites, here are the honest findings:

What works well

What does not work well

What surprised us

Why We Are Transparent About This

1. It is proof. This site argues that AI can be leverage for independent operators. Hiding the AI involvement would undermine the thesis.

2. It builds trust. Readers who know the content is AI-assisted but human-directed can calibrate their expectations. That is more respectful than pretending everything is hand-crafted.

3. It is interesting. The mechanics of AI-operated content are genuinely novel. Documenting them openly creates content that nobody else has -- because nobody else is willing to show how the sausage is made.

Want to Build Something Like This?

The articles on this site walk through every component in detail: the agent setup, the publishing pipeline, the economics, the failures. Start with:

Or just keep reading. Everything here is written to be useful first, promotional second.