How this works

A specification is a row, not a column

Every figure on this site is stored as its own record: the subject, the attribute, the value, the unit, the document it came from, and when it was extracted. Nothing is a blob of JSON attached to a model. That shape is what makes the rest of it possible.

Units are normalised, and the original is kept

One manufacturer publishes payload in kilograms, another in pounds. Both are stored as published, and both are converted to a canonical unit for the attribute. Comparison and filtering use the converted figure; the page shows you what the source actually said. A unit we do not recognise is parked for review rather than guessed at.

Sources disagree, and that is content

When a manufacturer says 35 kg and a review says 30, both rows exist. A view picks which to lead with — by how close the source is to the manufacturer, then by recency — and the page marks the figure disputed rather than quietly choosing.

What is written by a machine, and what is not

Specification values are extracted from documents, never generated. Model summaries are machine-written and labelled as such, everywhere they appear, including in the JSON API. Editorial is written from this database — from what changed, and what the numbers say — never by rewriting somebody else’s article.

Corrections

A wrong figure is a bug with an address: every value links to the document it came from, so a correction can point at the exact source. Where a document should not have been used at all, removing it removes the figures that rested on it.

DroidsBot, our crawler

The documents behind this site are read by a crawler that identifies itself as DroidsBot/0.1 (+https://droid.monster/about). It reads manufacturers’ product pages, sitemaps, datasheets and manuals — public pages only, never anything behind a login — so that each specification here can link to the document it came from.

It obeys robots.txt under the name DroidsBot, including Crawl-delay. It makes one request to a site at a time, at least two seconds apart, and backs off when a site answers 401, 403 or 429. If robots.txt cannot be read because of a server error, it treats the whole site as off limits.

What we publish is the figures taken from a document, each with a link back to it. The documents themselves are kept only as the record of where a figure came from, and are not republished. Product photos are credited to the page they appear on, and removed on request.

To stop it, add User-agent: DroidsBot with Disallow: / to your robots.txt.