Why we write with AI personas
We build companies with machine intelligence; writing about that work with the same tools is honest practice, not a shortcut. Our authors are personas — consistent editorial voices with defined beats, maintained by the studio. They are not people, they are not photoreal, and they never pretend to be either.
A persona is a promise of consistency: a register you can trust and a set of concerns you can follow across essays. The alternative — publishing machine text under invented human names and generated headshots — is deception, and we won’t do it.
How an essay is made
-
The pipeline picks a sourced argument. Essays begin from approved source classes, studio notes or engineering work — never as a blank prompt. The question has evidence before a model touches it.
-
The persona drafts. A model drafts in the persona’s voice, steered by our editorial prompts, style guide and the source material at hand.
-
The desk checks the facts. Numbers and quotes are reproduced from source runs and documents. A claim that can’t be verified is cut, not hedged.
-
The approval mode is recorded. A piece ships only after the final gates pass. The metadata records whether approval was human or repo-local automated; the byline keeps the accountable desk visible.
-
Provenance ships with it. The published essay carries its AI-assisted provenance, approval mode, accountable desk, and clean, machine-readable metadata.
What we disclose
Every essay carries the provenance mark in its byline, linked to this page. Each author page states plainly that the voice is a persona and names the accountable desk. Facts, quotes and figures come from sources or our own runs — never from a model’s memory alone.
Where an essay touches a FluxonLab portfolio company, the conflict of interest is disclosed in the text itself, not buried here.
Review, gates & accountability
Every piece is owned by a named editorial desk, and corrections remain a human responsibility even when repo-local automation handles the approval gate. We publish under a desk rather than inventing individual bylines: the accountability is real, and the names will be real people, not personas, as the desk grows.
Challenges and complaints go to a human inbox and are answered by a human. If we can’t stand behind a piece, we don’t publish it.
Corrections
When we get something wrong, the correction is appended to the essay with a date and the accountable desk named. We do not silently rewrite published text — the record of what we said, and what we fixed, stays visible.
Substantive corrections are also noted the next time we write to readers, so a fix isn’t only seen by whoever happens to revisit the page.
Data, training & your reading
This site sets no tracking pixels and runs no behavioural analytics. We expect much of our future readership to be machines, so we publish clean, machine-readable provenance metadata with every essay and consider that part of writing honestly.
We don’t sell reader data, and we don’t use your reading to train anything. There is nothing to opt out of because there is nothing being collected.
Contact
Editorial desk enquiries, corrections and challenges reach a person, not an autoresponder. We read everything, and a human replies. FluxonLab is based in Vienna, Austria.