The tests that stop this website from making things up
A site is easy to fill with claims nobody can check. The tests in this one's repository fail when a false shape comes back, and what they cannot catch.
Valentir Engineering Desk
A marketing site is easy to fill with claims nobody can check: a client logo, a round number, a promise about speed. This website's repository holds tests whose only job is to fail when a known kind of unprovable claim comes back. This note lists what they check, and what they cannot.
What they refuse#
- Invented or borrowed names. One test reads every shipped source file and fails if it names a client the studio does not have or a product that does not exist.
- A history we do not have. The studio was founded in 2026, so tests fail on a made-up timeline, on "projects shipped" counters, on a trust strip and on any median build time.
- A person's name. No individual is named on the public site, and every post is bylined by the engineering desk.
- A hand-typed count. The number of apps is not written anywhere. It is counted from the table of published installers, and a test fails if a number is written beside the word "apps" in the source.
- A hand-typed price. Every price on the site is read from the Pricing list. A test scans the source for rupee figures, and a property test checks that every price, payment split and warranty period in the copy is the list's.
- Two answers to one question. There is one reply time, held in one constant, and a test fails if it is written out a second way.
- A privacy page that denies what the site does. It cannot say there are no cookies while analytics tags run.
- A claim of signing. The product pages may say the installers are not code-signed. They may never say they are signed.
- Stale wording. A scanner covers everything the site ships, and the text the copy functions write, for phrases that a change to the price list on 29-10-2026 made false, such as an old payment split.
The numbers#
On 01-11-2026 the website repository held more than 700 test cases in more than 40 test files, found by searching for test declarations. The count moves every week, so treat it as a snapshot.
What they cannot catch#
A test can only fail on a shape someone thought of. It cannot tell whether a sentence is true. It also cannot see text that does not live in the code. Posts, case studies and testimonials arrive from the admin panel's database, so the site checks those at read time as well: a case study or testimonial that lacks consent, proof or a number's source is dropped, and a post with almost no text, or with a person's name as its byline, is not shown as it is.
A person still has to read the page. The posts on this site carry their own test as well: every post must name its evidence, and none may carry a percentage or a speed claim without a source.
Why publish this#
Because the claim "we do not make things up" is itself a claim. Showing the mechanism lets you judge it. Two of the rules above have notes of their own: why every price lives in one place and what a case study must have before we may publish it. Valentir Technologies LLP is a software studio in Noida, and the products page is where the thing that can be checked today lives.
Get our notes by email
We email our notes when we publish something worth your time.
Search on Google and want more notes like this in your results? Add Valentir as a preferred source on Google.
If you want to argue with any of this, that is what hello@valentir.in is for.
