Fourteen pages, and every one of them is a real file.
There is no database behind this and no server deciding what to send you. Every address below was written out as finished HTML at build time, so a visitor gets a complete page on the first response rather than a blank shell that fills itself in afterwards. Three of these boxes are templates, and the number on each is how many pages it produced this build. The fourteen are what a visitor can land on; the not-found page is there for the addresses that do not exist.
/blog/. Those addresses still answer, with a permanent redirect to the matching /news/ page, so anything already shared or bookmarked keeps working instead of quietly turning into a dead link. Redirects are cheap; broken links you never hear about are not.
What happens between saving a file and the page being live.
Five steps, all of them automatic. The important property is where the work happens: everything expensive is done once, here, before anything ships. By the time a visitor arrives there is no work left to do, which is why the pages open the way they do on a phone on a bad connection.
Reads Noble’s own YouTube feed and writes what it finds to a file. The videos on the site are whatever the channel published, not a list somebody has to remember to update.
Runs the whole site once and writes each page out as finished HTML, including the share cards. A page that cannot be built fails here, in front of us, rather than in front of a visitor.
Writes the copyright and authorship notice into every page it just produced, so the attribution cannot be forgotten on a page somebody added later.
Collects the finished pages into one folder and writes the redirect and file-type rules alongside them, so the rules that govern the site travel with it instead of living in a control panel.
The folder is copied to servers around the world. Each visitor is answered by whichever one is nearest to them, which is most of why the site is fast for someone in another state.
8 seconds on a laptop, roughly one of which is waiting on the YouTube feed. That number matters more than it sounds: a build fast enough to run on every change is a build people actually run, and a slow one quietly becomes a thing nobody does before shipping.
Three kinds of content, one place each.
Nothing on the site is typed into two places. Each kind of content has one home, and every page that shows it reads from that home, which is the difference between a site that stays true and a site that slowly stops matching itself.
Every heading, paragraph, statistic and testimonial on the site. Kept apart from the code so the wording can be changed without touching how a page works, and so a claim can be traced to one line rather than hunted for.
Taken from the channel’s own feed each time the site is built. Publish a video, rebuild, and it is on the site with its own page. Nobody transcribes a title or pastes a thumbnail.
The state outlines behind the coverage section, worked out once and saved. The page draws it directly instead of asking a mapping service, so there is no third party involved in a visitor loading it.
One more file sits in that folder and is deliberately wired to nothing: the record of which articles are still awaiting review. It is kept out of the site’s own code on purpose, because anything the code imports gets shipped to the browser, and a private note is not something to hand a visitor along with the page.