Ada Politika Publishing System

Our work with Ada Politika began in 2024 with a news site built on WordPress. Two years later the publication had outgrown it: the daily volume was more than an off-the-shelf stack could carry, and the newsroom's needs had become too particular to solve with plugins. So we did not refresh the site. We rewrote it.

The old site was taken over as it stood. Its archive, its pictures and its address structure moved across intact, old links still resolve through redirects, and nothing built up in search was lost.

The news flow runs itself

The system polls its sources on a schedule, reading RSS, Atom, JSON and XML, and parsing the page directly where it has to. Every incoming story is checked against what is already published, the pictures inside it are downloaded off the source's server onto the publication's own disk, and the markup is cleaned. The source is always credited in plain sight.

AI beside the editor, not instead of one

The AI does not rewrite the story. What it does is the repetitive judgement that eats an editor's day: drawing a standfirst, proposing an SEO title and description, suggesting tags and a section, scoring how much a story matters locally and how much of a lead it is, catching stories that say the same thing in different words, writing the social copy, describing pictures for screen readers, and sifting comments before a person sees them. The monthly budget is set in the panel, and when it runs out publishing does not stop — only the enrichment does.

The front page can arrange itself

The headline boxes can be filled automatically from a news-value score. Each box is configured on its own: filled by hand, chosen by the system, or drawn only from one section. How long a story may hold a headline is set in the panel. A story an editor placed is never moved by the machine, and a box is never left empty.

Social distribution

Published stories go out to Facebook, Instagram and Telegram on their own. A story with no photograph gets a cover drawn for it before it reaches Instagram. Posts are queued and spaced apart, failures are retried according to what went wrong, and every one of them is on record in the panel.

Everything that is not a news story

Exchange and gold rates, a crypto listing, the weather, prayer times, duty pharmacies, newspaper front pages, football fixtures and the league table are all fetched on a schedule and given their own pages. Daily, weekly and monthly horoscopes publish themselves.

Analytics without personal data

The site counts its own readers and does it without keeping anything personal: no IP address is ever stored, a visitor's identity is derived with a salt that changes every night, no cookies are used, and a browser that asks not to be tracked is not recorded at all. Raw rows are deleted after a set period and only the summaries remain.

For search engines

Canonical addresses, robots directives, sitemaps, a Google News sitemap, NewsArticle and organisation markup, share cards and redirect management are all part of the system. Addresses that answer 404 are logged and can be turned into a redirect from the panel in one click.

Behind the scenes

Roles and permissions, two-step verification, an audit trail, system logs, job queues, source health screens, cache control, maintenance mode and automated backups. Every fixed string and every fixed image on the site is editable from the panel; none of it is written into the code.

Performance

An independent PageSpeed Insights run scored 98 on mobile and 100 on desktop, with zero layout shift and zero main-thread blocking. The system runs on shared hosting.