Skip to main content

Editorial

How this site is written and checked

Every article on Black Whole describes work we did ourselves: builds we ran, exports we inspected, failures we had to diagnose. This page explains who writes the site, what testing stands behind each claim, how AI assistance is used, and how to get something corrected.

Last updated 1 November 2026

Who writes the articles

The blog and guides are written by Nirpesh, the founder of Black Whole. The product is built by the same person who writes about it, which is why the articles talk about specific exports, bundle sizes, build logs and error messages instead of abstract advice.

Over the past year that has meant shipping a website builder that plans a project, writes real files, verifies its own output and publishes, a game agent that ships playable browser games, and a general agent that works from project files. The failures in that work, not the marketing, are what the articles document. You can read more about the company on the about page.

What testing stands behind a published claim

An article is only published once the thing it describes has been run on a real machine. When we quote a number, it came from a command we executed: file sizes from an actual export directory, exit codes from an actual build, the exact wording of an error from the terminal that produced it.

Where a result depends on the environment, we say so. A WebGL export that works on a normal browser cannot be visually verified inside a sandbox without WebGL2, and we write that limitation down rather than implying a screenshot was taken. Where something is a judgement rather than a measurement, it is written as a judgement.

Rules we hold ourselves to

  • No invented statistics, survey figures, study dates or customer quotes. If a number was not measured, it is not written.
  • No fabricated case studies. Examples come from our own builds or are clearly described as hypothetical.
  • No claim that a platform will approve, rank or pay out, because nobody outside that platform can promise it.
  • Steps are written in the order they were actually performed, including the parts that failed first.
  • Product limitations are stated in the same article that describes the feature, not buried in a footnote.

How AI assistance is used

Black Whole is an AI product, so it would be odd to pretend AI plays no part in the writing. Drafts are sometimes assembled with the help of AI tooling. Every article is then checked against the source of the claim: the terminal output, the file listing, the build log, or the platform screen it came from. Anything that cannot be checked is cut.

The rule we apply is simple. An AI may help us phrase a paragraph, but it does not get to invent a fact, and it does not get to make a number up to make a paragraph read better. When an article is heavily rewritten by hand after drafting, that is normal and not something we consider worth flagging.

Corrections and updates

Articles carry an updated date. It changes when the substance changes, not when the page is reformatted. If a product change makes an article wrong, the article is corrected or marked as historical rather than left to mislead readers who find it later.

If you find an error, send it through the contact page with the URL and the sentence you believe is wrong. Corrections with a specific source get priority. We would rather be corrected than be quoted.

Advertising and independence

Black Whole is a commercial product. Some pages carry advertising, and the site may earn from ads shown next to editorial content. Sponsorship never buys a favourable review, and no advertiser sees an article before it is published.

Comparisons on this site include products we compete with. Where we have tested an alternative, the article says what was tested and when. Where we have not, we say that too. Our advertising policy covers how ad placements are handled, and the privacy policy covers what is collected when ads load.

Topics this site covers

Black Whole writes about a narrow set of things: building websites with AI, writing briefs that models can act on, verifying generated code before it ships, running agents against real projects, exporting browser games to the web, and the practical details of hosting what comes out. Articles outside that focus are not published here, because a site that writes about everything is usually trusted for nothing.