What Parameter is

Parameter is a technology publication covering artificial intelligence for readers in the US, Europe and internationally. We do not claim to conduct original on-the-ground reporting unless a story explicitly says so. What we do is read widely across primary and secondary sources, establish what is actually confirmed, and explain it clearly with the sources attached.

How a story gets published

Every story passes through the same pipeline, and each stage can stop it.

  1. Discovery. We monitor official company blogs, press releases, regulator publications, research feeds and technology newsrooms through their published RSS feeds and APIs. We respect robots.txt, rate limits and terms of service. We do not scrape sites that do not offer a feed for this purpose.
  2. Deduplication and clustering. Reports about the same underlying event are grouped into a single story. Ten outlets covering one announcement produce one Parameter story, not ten.
  3. Fact extraction. Individual factual claims are extracted from each source and tracked separately, along with which source asserted them.
  4. Verification. Claims are marked confirmed when stated by the primary source (the company, lab or regulator making the announcement) or corroborated by two independent sources; reported when they come from a single credible secondary source; and disputed when sources disagree. Disputed claims are either presented as disputed, with the disagreement described, or omitted.
  5. Writing. We write an original article from the verified claim set — never a rewrite of a single source’s article.
  6. Quality control. Automated checks reject or flag drafts for insufficient sourcing, unsupported numbers, missing primary sources, excessive similarity to any single source, misleading headlines and other failures listed below.
  7. Editorial review. A draft that clears every automated check is marked approved and enters the newsroom queue. Approved is not published: it means the software found nothing wrong, which is a different and much weaker claim than a person deciding the story is worth running.
  8. Publication. An editor reads the draft and publishes it. Nothing on this site was put there by an automated process. Our publishing software has no route to the public site at all — the highest status it can assign is approved, and the change to published can only be made by a signed-in human being, whose name is recorded against it.

We would rather publish nothing on a quiet day than fill the front page with whatever scored highest. An empty queue is a normal outcome, not a fault.

Where AI assists, and where it does not

We use AI tools, and we would rather tell you exactly how than hide behind a generic disclosure. AI assists with:

  • monitoring and triaging incoming source material
  • grouping reports about the same event
  • extracting factual claims and comparing them across sources
  • classifying stories into sections and identifying entities
  • drafting article text from the verified claim set
  • generating headline, summary and metadata options

AI does not decide what is true. Specifically:

  • Every factual claim must trace to a source we can link to. A claim that cannot be traced is marked uncertain or removed.
  • We do not publish AI-generated quotes. A quotation appears only when it appears in a cited source, attributed to that source.
  • We do not publish statistics, prices, model specifications, dates or funding figures that are not in a cited source.
  • Articles carrying AI assistance say so. The notice appears at the foot of every affected story.

AI also does not decide what is published. A draft produced with AI assistance is material for an editor, and it reaches readers only when a person has read it and published it. Where a draft could not be written properly — no model available, or too little verified material to work from — the pipeline produces a working summary for the newsroom rather than an article, and that summary is not publishable: our software refuses to put it on the public site, and refuses to let an editor publish it either, until it has been written.

Editorial responsibility for everything published here rests with Parameter, not with a model or a vendor.

What we reject

A draft does not publish if it:

  • rests on a single source with no primary confirmation for a significant claim
  • duplicates a story we have already covered rather than updating it
  • contains a claim that no cited source supports
  • contains a quotation not found in a cited source
  • reproduces too much of any one source’s wording
  • is too thin to be worth a reader’s time
  • carries a headline the article does not support
  • repeats keywords in a way that serves search engines rather than readers
  • contains dates or numbers that contradict the sources
  • reuses a source’s headline as its own
  • claims work we did not do — “we tested”, “our benchmark”, “we interviewed” — where the work was somebody else’s
  • contains language describing its own editorial status rather than the story

The last three are recent, and they are here because we got them wrong. In September 2026 seven articles went to this site carrying another publication’s headline, another publication’s opening paragraphs, and a notice saying they had not been edited. They were removed and the software was changed so that each of those three faults now blocks publication outright. We have written this down rather than quietly fixing it because a policy page that only lists a publication’s intentions is worth less than one that admits what it has had to correct.

Sources and attribution

Every story lists its sources, grouped by whether they are the primary source or additional reporting. We link to the original rather than reproducing it, and we do not present other newsrooms’ reporting as our own. The full list of feeds we monitor is published on our sources page.

Reporting, analysis and opinion

Nearly everything we publish is reporting: an account of something that happened, built from sources we name. Where a piece argues rather than reports — a judgement about what something means, a prediction, an assessment of whether a claim holds up — it is labelled as analysis at the top and written in a voice that makes clear it is our reading rather than a fact. We do not run unsigned opinion, and we do not let an argument travel under a reporting headline.

The “Why it matters” section on an article is interpretation, and is meant to be read as ours. Everything above it is sourced.

Copyright and fair use

We read other publications through the feeds they publish for that purpose, and we quote from them briefly, with attribution and a link, to report what they have said. We do not republish articles, we do not reproduce them at length, and we do not use another newsroom’s work in a way that substitutes for reading it — which is why our source lists link out rather than excerpt. Our software measures how much of any single source a draft reproduces and refuses to publish a draft that leans too heavily on one, and separately refuses any draft that reuses a source’s headline.

We do not copy photographs from other news websites; see Images below. If you are a publisher and believe we have taken more of your work than is fair, write to us through the contact page and we will take it down or cut it back while we look at it, rather than after.

Corrections and updates

News changes. When new information materially changes a story, we update the existing article rather than publishing a second one, and we timestamp the update. When we get something wrong, we correct it, label the correction on the article, and list it on our corrections page. We do not quietly edit errors away.

If you believe something here is wrong, tell us through the contact page. Include the story and what specifically is inaccurate; we will check it against the sources.

Images

We do not copy photographs from other news websites. Article images are press-kit material used within its stated terms, licensed stock, public-domain material, or original illustrations produced by us. The licence and provenance of every image is recorded.

Independence and conflicts

Sponsored content, if we ever publish it, will be labelled as sponsored at the top of the article and excluded from editorial sections. Affiliate links, if used, will be disclosed on the page where they appear. Coverage decisions are not for sale, and advertisers get no advance sight of stories.

Ownership and contact

Parameter is an independent publication. It takes no funding from the companies it covers. Editorial enquiries, corrections and complaints: contact us.