Editorial policy

Who is responsible

Pijus holds editorial responsibility for everything published on SciTech Journal. Nothing appears on this site without a person reading it first and deciding it is good enough to publish. If an article is wrong, that is a person’s mistake, not a machine’s.

Where articles come from

There are two routes. Some articles are submitted by contributors — anyone with a Google account can write one. Others start in-house, where an editor picks a topic and uses an AI research assistant to gather sources and produce a first draft.

Articles from the second route say so in the byline: AI-drafted, fact-checked and edited by Pijus. That label links back to this page. Contributor articles carry the contributor’s name and no such label.

What the AI actually does

On an in-house article, the assistant gathers current sources on the topic, writes the first draft of the explainer, and proposes the diagrams. It writes the words. We are not going to describe that as light assistance, because it isn’t.

What it does not do is decide what gets published. It cannot publish. The draft lands in a review queue and stays there until a person clears it.

What the review involves

Review here means reading the draft against its sources, not skimming it for typos. Before anything is published the editor:

• checks every factual claim that matters against the cited source, and removes claims that cannot be traced to one;
• checks that each source is real, current and actually says what the draft says it says;
• checks the diagrams describe the real mechanism, since a plausible-looking wrong diagram is worse than no diagram;
• rewrites whatever is vague, padded or hedged into something concrete;
• cuts the article entirely if it turns out to have nothing worth saying.

Drafts do get killed at this stage. That is the point of having the stage.

What we do not do

We do not publish unedited AI output. We do not set volume targets or publish on a quota, because the moment the schedule outruns the checking, the checking is what gives. We would rather publish one article a week that is right than five that are probably right.

Sources

Articles built from research list their sources at the foot of the page, linked, so you can go and read them yourself. If an article makes a claim you want to verify and the source is not there, that is a fault — tell us.

Diagrams

Technical diagrams are drawn as code, not generated as images. A generated picture of a protocol tends to look right and mean nothing. Drawing it as code forces the diagram to commit to how the thing actually works, and lets us fix it when it is wrong.

Corrections

If you find an error, tell us and we will fix it. Substantive corrections — anything that changes what the article means — are made in the text and the article’s updated date changes with them. We do not quietly rewrite articles to cover a mistake.

Why this page exists

Partly so you know what you are reading. Partly because under the EU AI Act, published AI-generated text is exempt from a blanket disclosure requirement only where it has had genuine human review and a named person is responsible for it. We would rather write down what our process is than claim one.

Questions about any of this, or a correction to report? See the about page.