NAKKAMA

Editorial Standards

How this magazine is made, by whom, with which tools, and what we refuse to do.

Nakkama writes about operational AI. Publishing that without saying how the site itself is produced would be the first dishonesty. This page exists to avoid it.

Who writes

Nakkama is written, edited and published by one person: Julien Abbou. No newsroom, no committee, no agency. Every text carries his signature.

Nakkama is a point of view. That of an operator who actually builds the tools he writes about, and who reads the trajectory of technology through critical and cyberpunk thought.

Editorial responsibility for everything published here is whole and personal.

How an article is made

The chain, without euphemism.

1. The subject. A precise thesis, taken by a human from a reservoir of subjects built by hand. Never a generated theme, never a subject kept because it performs.

2. The lens. A real thinker whose concept directly illuminates that thesis. Verified before being used, never name dropped as decoration.

3. Assisted writing. Language models are involved, inside an internal tool built specifically for this magazine and loaded with a hand written voice guide: one thesis per article, a tone, a list of prohibitions, a rule of truth. The model produces material. It does not produce an article.

4. Refinement. No text is published as it leaves a model. Every article goes through a human pass that cuts, arbitrates and rewrites, and through a reference by reference check.

5. Publication. Signature, date, responsibility.

In one sentence: AI is a tool in the chain, not the author. The thesis, the selection, the judgement and the final read are human.

The rule that comes before style

No invented fact, figure, date, quotation, author or book. No sentence attributed to someone who never said it. No thesis lent to a thinker who never argued it. No link to an article on this site that does not exist.

When in doubt, we leave it out. A cautious true statement beats a brilliant false one.

This rule outranks effect, style, and the urge to publish.

Images: the artwork and the fault

Every cover is a public domain artwork, drawn from museum open access collections (The Met, Art Institute of Chicago, Rijksmuseum, Wikimedia Commons), crossed by a computer generated analogue signal fault: a rolling sync bar, split chromatic fringes, dropout, snow.

It is not a filter. It is the magazine's argument made visible: the classical canon crossed by a breakdown, the original and its corrupted simulacrum.

Coherence with the article is not decorative, it is calculated. The artwork is searched from the thesis of the text, not from its vague theme: a vanitas for the end of something, a court scene for an article about power. The type of fault and its intensity are derived from the tone of the article: dark, middle or light. A grave text receives a collapsing signal, cold and desaturated. An ironic one receives a vivid, coloured glitch.

The alteration is declared, never hidden. It carries a unique number, and the caption always credits the source artwork, its artist, its year and its museum.

No image on this site documents a real event. These are covers, not press photographs. None depicts a real identifiable person.

Sound: the same signal, in another material

Each article has a sonic climate, never a soundtrack. Sound does not illustrate the text, it extends it. It is what you would hear approaching a broadcast with poor reception.

It follows the same coherence rule as the image, and inherits exactly the same tone: one signal, two senses. The climate is built from the central image of the article, not from an interchangeable library of ambiences. A text about the informal circuits of an organisation receives a muffled open plan office and footsteps in a corridor. A text about the hand that leaves a trace receives a workshop, sanding, slow gestures.

The material comes from real recordings under CC0 or CC-BY licence, degraded afterwards, never from a synthesis that imitates. Every sound is credited with its source and its licence.

The podcast

Fréquence Nakkama uses synthetic voices. None seeks to imitate a real person, and none is a clone of an existing voice.

The host introduces himself on air as the digital replica of Nakkama's founder. The formula is spoken in every episode. It is not a stylistic effect: it is the distinction between the author and his replica, and that distinction is the whole subject.

Episodes derive from published articles, never from a parallel corpus. The source article is always linked.

Corrections

A reported error is corrected.

  • Substantive error: corrected with a dated note at the foot of the article. A text is never silently altered on a point that changes its meaning.
  • Typo, dead link, typographic slip: corrected without a note.
  • Withdrawn article: the address stays alive, with an explanation of the withdrawal.

To report an error: contact@nakkama.com

Independence

Nakkama is an independent, personal project. It commits no organisation, no employer, no client.

It is an experiment, assumed as such, in substance and in form: actually building the production chain it writes about, in order to test a thesis rather than assert it, and to hold a concrete case for teaching. The articles are the result of that experiment, not a service.

Opinions expressed are the author's alone.

Funding

None. Nakkama sells no advertising space, publishes no sponsored content, runs no affiliate links, and receives no consideration for anything published here.

No subject has ever been kept because it paid. None has ever been dropped for the same reason.

Languages

Nakkama publishes in French, English, Spanish and Chinese. French is the language of origin. The other versions are translations assisted by language model, reviewed before publication. Where two versions diverge, the French version prevails.

Last updated: 14 August 2026. This page changes when the production chain changes, and the change is dated.

← → previous / next · ↑ ↓ random
·

···