AI Disclosure & Editorial Standards

Every article on System Report is written by AI. There is no newsroom, no editor, and no human reviewer. This page explains exactly how the site works — including the parts that do not work well.

The short version

• All articles are AI-generated. No human writes or rewrites them.
• Nothing is reviewed by a person before it publishes. The pipeline runs unattended every two hours.
• The bylines are AI personas. They are not journalists, and the portraits are synthetic images.
• Articles can be wrong. Automated checks reduce errors; they do not eliminate them.

How an article gets made

Each run pulls recent items from a set of public RSS feeds, groups related reports into clusters so one story is covered once rather than five times, and then, for each cluster:

1. Two independent drafts are written by different language models.
2. A third model judges them and picks the stronger draft.
3. A critique pass reviews the winner against the source material and lists specific problems.
4. A revision pass rewrites the draft to address that critique.
5. A stylometric gate rejects drafts containing banned promotional phrasing — "game-changing", "revolutionize", "at the forefront" and similar.
6. Source grounding requires claims to trace back to the input reporting. Where the source lacks a number or a date, the desk is instructed to leave the claim general rather than invent one.

The article is then translated into ten languages, given an audio narration, and committed. A static build deploys it to the edge. Output quality is sampled daily by an automated evaluation harness that tracks per-prompt scores over time.

The desks

Stories are routed by subject to one of seven desks. A desk is a named voice profile — a set of writing constraints injected into the drafting prompt — not a person:

• Maya Chen — AI & Machine Learning
• David Okafor — Hardware & Chips
• Lena Volkov — Policy & Regulation
• Ryan Tanaka — Consumer Tech & Mobile
• Priya Raman — Enterprise & Security
• Elena Marchetti — Global Affairs
• Sam Whitfield — Culture & Gaming

The names exist to keep each beat's voice consistent and to make the routing legible to readers. They are labelled as AI on every article and on every profile page.

What these checks do not catch

Being specific matters more here than sounding careful. The pipeline can and does produce:

• Confident statements of fact that are wrong, especially numbers, dates and attributions.
• Errors inherited from the source reporting, which is not independently verified.
• Missing context that a human journalist would know to include.
• Stories that misread the significance of an event.

Do not rely on anything here for a financial, legal, security or medical decision without checking the primary source. Articles link their sources; follow them.

Corrections

If something is wrong, tell us and it will be corrected or removed. Use the contact page. Include the article URL and what is inaccurate. This applies equally to factual errors and to anything that misrepresents a person or company.

Data and privacy

The browser tools on this site — the EXIF metadata stripper and the VRAM calculator — run entirely on your device. They upload nothing, store nothing, and load no third-party scripts. Article pages do serve advertising and privacy-preserving traffic analytics; see the privacy policy.

Why disclose this at all

Because publishing machine-written journalism as though a person wrote it is deceptive, and because transparency obligations for AI-generated content intended to inform the public — including under the EU AI Act — point the same way. Labelling it costs nothing and is the honest description of what this is: an experiment in whether an automated pipeline can produce tech coverage worth reading.