Changelog
WORK LOG
What we shipped, what we measured, and what we got wrong and then fixed. With the date and a link to what changed. The date is the day the change entered the code, not the day it was announced. Our mistakes are here too, because a site that hides them hides other things as well.
The site has not launched on its own domain yet: the company is being registered. A launch entry will appear here when the launch actually happens, not before.
We removed the address of a practice we had no right to name
The work page and the Niš page showed the address of a practice whose site we had measured as found. Our own rule says the name of a practice with a bad site is never published anywhere, and an address identifies exactly what a name does. The addresses are gone. The check meant to prevent this existed and passed: it tested one hardcoded record instead of the rule, so the rest of the list went unchecked.
What it means for you If we measure your site as an example of the current state of things, neither your company’s name nor its address appears on our site.
We published a better number than the real one on the English page
The board claimed the largest content painted in 1.59 seconds on the English page. Measurement said 2.59, above our own threshold. The cause was the hero subtitle fading in, so the measurement waited for the animation. Fixed, and it now measures 1.58 s. The Serbian page passed at 1.70 s, which was luck and not merit.
What it means for you A speed number we show you is measured, and when it is worse than the threshold, that is what it says.
The published JavaScript figure lagged six days behind the real one
On a site that sells measurement we published 160.2 KB while the reality was 165.7. The test only asked whether the number was under the threshold, not whether it was true, so the drift passed. The difference ran in our favour, which is the worst possible direction. Since then the published number must track the one from the build, or the suite fails.
What it means for you The number on the Status page has to match the latest build; when they drift apart, a check fails before the change goes out.
Three pages showed the same borrowed work as proof
The pages for clinics, salons and those without a site all showed the same case, so anyone opening two concluded we had no examples. The conclusion was correct. Borrowed proof from a neighbouring trade is gone, and those pages now say there is no proof for that build yet.
What it means for you When we do not have an example for your line of business yet, the page says so instead of showing you someone else’s.
Accessibility statement
Nine checks in a table, with tool and date. Five manual ones are not done yet, and the page says so above the table.
What it means for you Before you rely on the site, you know what has been checked on it and what has not.
Eight pieces in three layers
Price, comparison, guide. Each carries a number of our own or a source you can open: a test opens that source and looks for the number in it.
What it means for you You can check every number in the articles yourself, at the source listed next to it.
A local page for Niš
What “local” concretely means, and one measured site in Niš. No map, because a map with four dots is not data.
What it means for you If you are from Niš, you see what working with a local studio means for you, without general promises.
The public handbook
Method, terms of the offer, what we turn down, what gets handed over, and how we use Claude Code. Five chapters, four tables.
What it means for you You can read the rules we work by before a call, and point to them afterwards.
Five steps from enquiry to care
A track that slides as the page scrolls, without a single byte of JavaScript. Every step has an output that stays with the client.
What it means for you You know in advance which step comes next and what you keep after each one.
Case studies with measured figures
Four published. Work without a measured figure is not published: an empty shelf beats a picture with nothing behind it.
What it means for you The examples you see carry a measured number; we do not show a site without one as proof.
A questionnaire that builds the quote
A few short questions, then a price: before we ask for a single detail about you. Where the price list does not cover something, the result is a range with both ends from the list.
What it means for you You see the price before you give us your name, phone number or e-mail.
Public prices
The whole price list on one page, the same for everyone. The price on screen and the price in structured data come from the same file.
What it means for you The price you see is the same for everyone and does not depend on who is asking.
Call booking
A slot in your own time zone, in four steps or fewer, with no account to create.
What it means for you You pick the time of the call yourself, in your own time zone, without opening an account.