Editorial Process
How summaries reach Interestana, the sources we trust, and the safeguards we run in between. Last updated 8 June 2026.
How an article reaches Interestana
Every published summary on Interestana follows a five-stage pipeline. First, one of twenty specialised agents — each scoped to a category such as Artificial Intelligence, Cybersecurity, or Economics — fetches the latest items from a hand-curated set of weighted RSS feeds. Second, items older than forty-eight hours and items already covered by a previous run are filtered out. Third, an AI model (Google Gemini as the primary, OpenAI GPT-4o-mini as the fallback) produces a 300 to 500 word factual summary in the form of three to five plain-prose paragraphs. Fourth, the summary is parsed and validated against our taxonomy — labels are normalised, the category is locked to the agent's scope, and the read time is bounded between one and fifteen minutes. Fifth, the article is stored in our MySQL database with a stable canonical slug, indexed in the sitemap, and made available through topic pages, the RSS feed, and the homepage.
Sources we rely on
Each agent's sources are chosen for editorial reputation, publication cadence, and topical specificity. Examples include MIT Technology Review and VentureBeat for AI; The Verge and Ars Technica for consumer technology; Bloomberg, the Financial Times, and Reuters for economics; The Economist for global affairs. Each source is assigned a weight between one and five that biases our deduplication and ordering — higher-weighted publications win ties and surface earlier. We update the source list when feeds break, when new authoritative outlets emerge, or when an existing source persistently fails our quality bar.
What our summaries are — and what they are not
Interestana summaries are reference snapshots, not original reporting. They distil the core facts — who, what, when, where, and the most consequential implications — from the source article so a reader can decide quickly whether the underlying story is relevant to them. Every summary on Interestana is accompanied by a prominent link to the original publication. We do not paraphrase to disguise borrowing; we cite. We do not attribute opinions to ourselves that belong to the source; we attribute. And we never publish a summary if our pipeline cannot reliably produce a coherent and accurate one — failed runs are logged, not shipped.
Editorial review
Our pipeline is AI-drafted and human-overseen. The publisher reviews flagged outputs, samples published summaries each week, and audits the agent taxonomy on a rolling basis. When an automated summary misrepresents a source, omits a critical caveat, or contains a verifiable error, the published article is corrected in place. Corrections are appended to the article with a short note explaining what changed and when. We do not silently rewrite history.
Corrections and takedowns
If you spot an inaccuracy, please contact us through the contact page and reference the article URL. We aim to acknowledge correction requests within twenty-four hours and to publish or reject the change within seventy-two hours, depending on the complexity of verification. Publishers who believe their work is being summarised in a way that exceeds fair use, or whose RSS feeds we are using in a way they object to, may also request removal through the same channel.
Independence and conflict of interest
Interestana publishes no paid editorial. We accept no money in exchange for coverage, source weighting, or favourable summaries. When a topic we cover overlaps with the interests of the founder, his consulting clients, or the Interestana Agency arm, we disclose that overlap on the article. Our agency clients have no influence over editorial coverage decisions or source selection.
Privacy of subjects and readers
We do not collect personal data from readers beyond aggregate analytics provided by Google Analytics 4. We never reproduce personally identifying information from sources beyond what those sources have already made public. When summarising stories involving private individuals, we err on the side of restraint and prefer to link rather than restate.
Open source models and citation
Interestana is built to be referenceable. We publish a structured sitemap covering every article, expose article facts via Schema.org NewsArticle structured data, allow AI crawlers (GPTBot, ClaudeBot, Google-Extended, and others) in our robots.txt, and link out to source publications by name. Our goal is to make every summary on Interestana a useful and traceable citation — both for human readers and for AI systems that draw on the open web.