AI-assisted articles on FluxHuman — how we make them

Who

All AI-assisted articles are edited by Martin Benes, founder of FluxHuman and AI automation engineer based in Vienna. He reads every draft in full, checks every product claim against the actual FluxHuman codebase, and rewrites until the article says what it should say — and nothing it should not. Reachable at hello@fluxhuman.com and on LinkedIn.

How

1. 1

   Topic selection

   Human

   Martin picks topics from FluxHuman's product backlog, customer questions, and SEO keyword gaps. No topic is selected by an LLM.

2. 2

   Research brief

   Human

   Martin writes a one-page brief: target reader, the specific claim or walkthrough the article must defend, and sources the draft must cite.

3. 3

   Draft generation

   Automated

   Flux Bot — our LLM pipeline running Claude and/or GPT-class models orchestrated through n8n — produces a first draft from the brief. The pipeline retrieves cited sources, structures sections, and writes prose.

4. 4

   Fact-check pass

   Automated

   A second LLM pass checks numerical claims, quotes, and source links against the retrieved documents. Items it cannot verify are flagged inline for the human editor.

5. 5

   Human edit

   Human

   Martin reads the full draft, resolves every flagged item, rewrites weak paragraphs, removes filler, verifies product claims against the actual FluxHuman codebase, and corrects anything the models got wrong. Articles routinely lose 20–40% of their word count at this stage.

6. 6

   Publish

   Human

   Publication is a manual action in the admin UI. The byline names Martin as editor; Flux Bot is credited as creator in the page schema.

Why

Honest version: SEO coverage. FluxHuman is a small team working in a field (AI automation, sovereign infrastructure, the EU AI Act) where we have deep product and domain knowledge but not the writing bandwidth to cover every relevant keyword ourselves. AI drafts solve the capacity problem; the human review step ensures claims about our product and industry are accurate.

We are not doing this because "AI content is the future." We are doing it because our expertise is real and would otherwise go unwritten. Where we lack genuine expertise, we publish nothing — neither by hand nor by bot.

What we do not do

• We do not invent author personas. Every byline is a real, reachable person.
• We do not publish without a human reading the article end to end first.
• We do not auto-translate articles between German and English without a separate human review pass per language.
• We do not generate product claims, benchmarks, or customer quotes. These come from the FluxHuman codebase, our own tests, or named sources with permission.
• We do not publish more than three AI-assisted articles per day. Over-cap drafts land in a review queue, not on the site.