One feed item is one borrowed story
A tool that takes a single RSS item and asks a model to say it differently hands back a variant of someone else's work. When the same story runs across ten outlets, you get ten variants and no advantage.
For newsrooms
Ten sites are rewriting the same story. You run the one piece that takes what matters from each of them.
SemanticHub gathers every report on the event, pulls the facts that count out of each one, searches out the context they are missing and corroborates it in further sources. It writes in your structure and your voice, end to end, on its own. Your editor is left with one move: publish or bin it.
No onboarding project, no API key of your own, no sales call.
Rewriting does not build a publication.
Readers have no reason to come back
A site stitched together from repetition gives nobody a reason to pick it. The value starts where someone compares the reports, catches the contradictions and says what follows from them.
The liability stays with the publisher
Your masthead answers for what you publish, not your vendor. So every draft has to carry its sources, and the call to publish has to sit with an editor.
From sources to a finished draft.
- 01
You pick the sources
RSS feeds, site addresses, sitemaps. Add the outlets your desk reads daily; the rest of the catalogue is already there.
- 02
The system assembles the topic
Reports covering the same event land in one cluster, decided by what they say rather than by matching headlines. Ten write-ups become one item to handle instead of ten.
- 03
A filter holds the beat
You describe, in your own words, what counts as a story for your title and what does not. Clusters outside that brief drop out before anyone spends a minute on them.
- 04
The draft waits for an editor
The text is written from the whole cluster, in the structure you set for the goal, with links to every source. It moves only once approved — by webhook, by email or through a CMS plugin.
Nothing reaches the site without an editor.
A queue, not autopilot
Every draft stops in the queue by default. Hands-off delivery is something you switch on yourself, per goal, once you trust the setup.
You can see what it was built from
Each draft carries the full set of source articles with links and dates. Your editor checks the fact at the source instead of taking the model's word for it.
Test mode to begin with
A new goal delivers to a dead-end address, so you can watch the first runs without anything landing on your live site.
No API key of your own, no second invoice.
The model is in the subscription
You don't open an account with a model provider and you don't get a separate bill for tokens. The number of drafts sits in your plan and is written out in the pricing.
Servers and database in Poland
The application and its database run in Poland, the data processing agreement is available in Polish and English, and the full subprocessor list is in the legal documents.
Polish is a first-class language
The interface, the docs and the support are written in Polish, and the filter and drafts work in Polish natively rather than in translation.
Before you sign up.
- How is this different from tools that rewrite RSS articles?
- The unit of work is a topic, not a single borrowed article. Every report on the same event is gathered into one cluster first, and the draft is written from the whole set, with a link to each source.
- Can anything publish without the newsroom agreeing?
- No. Drafts wait in a queue and delivery starts on approval. Automatic dispatch is a separate per-goal setting and it is off by default.
- Where do the sources come from, and can I add my own?
- The catalogue of outlets is shared, and you add your own while creating a goal: through search, by site address, or by pointing at an RSS feed. A goal can run entirely on titles you chose.
- Will the text end up a duplicate of another site's article?
- It is written from several reports at once and in the structure you set for the goal, so it isn't a reworded single input. The publishing decision still belongs to an editor who can see every source.
- Do I pay for the model separately?
- No. The model cost is inside the plan, you don't connect a key of your own and you don't settle up with a model provider. The pricing page says how many drafts each plan covers.
- Does it work with WordPress?
- Yes. Finished drafts arrive through a plugin that creates post drafts in your CMS, through your own webhook endpoint, or as plain email to the desk.
Start with a single goal.
Create an account, point it at a few sources and see what the first cluster looks like. No onboarding project, no sales call.