The News

Started June 2026, ongoing, currently live Link: thenews.masteradamo.uk

Every day, like everyone else, I read the news, and what I see, which must be the same thing everyone else sees, is a distended organ. Instead of a presentation of the world-as-it-is, I see an arena for all the characters out there with an ideology to pitch, a fortune to make, a bridge to sell or cross or burn, to come out and make a spectacle of themselves. The problem with the news as I see it isn’t the content itself; the problem is the way that the news has become an instrument of its own content. Almost every story ever told is about people, but with the advent of algorithm-driven broadcasting, what a story is about gets separated from what the story does, and what I see these news stories doing is telling us what to think and how to act, like fairytales for a voyeuristic public.

And so I read the news, and I see all the slop, all the spin, all the dross that emerges from the murky processes of politics and economy, all the manufactured crises, all the sponsored hit pieces, all the glowing exposés, all the sects and partisans and cabals with a PR department; I see every fake smile, every fake handshake, every fake hairdo, every deep fake, every altered photo, every fixed game, every false flag, every false economist; I see all you dog whistlers, all you cat callers, all you let-it-flow bros, all you aircon salespeople, all you techno-solutionists, all you neo-feudalists with your moats and your defences; I see all you virtue signallers, all you exclusionary progressives, all you single-member clubs and the empty sets you leave behind, all you conspiring theorists – I read the news and I see all of you, and it strikes me that the problem with our planet at the moment isn’t the people we’re led by; it’s that we let people lead us in the first place.

And then I had an idea: what if the news actually wasn’t about people, or at least not about specific people? What would happen if we stripped the all the names and associated personalities out of the news and just presented it as a catalogue of incidents between a bunch of anonymous characters? The leaders of nations would be reduced to the status of bureaucrats, company leaders and industrialists rendered as administrators and technicians. Corporations themselves could likewise be debranded, recast as generic stage props in the background of world events.

Could the news redone as a kind of pablum, a rolling and perfunctory report on the political, financial, and cultural weather, alleviate some of the anxiety that comes from a news cycle treated as though it is a loudspeaker for every despot and crackpot who would like to control the rest of the world? The point of The News is to offer a glimpse of what this might feel like.

What it is

The News is a website that gets updated once every six hours with ten of the top news stories in the world. The system that automatically manages the site works by grabbing live news feeds from a range of news sources, clustering stories from these feeds into topics, and then rewriting the most widely reported topics into depersonalised news stories. The rewrites are done by an LLM, which is prompted to remove any mentions of specific people and, where possible, corporations, and instead use more generic terms to refer to these individuals and organisations. So, for instance, a story about British politics might make reference to “the Prime Minister of the United Kingdom” or possibly employ a metonym like “Downing Street”, rather than refer to that individual by name.

The result of this depersonalisation is a version of the news that can feel very subdued and maybe also a little naive. It is certainly lacking in the knowingness with which the news is presented in mainstream media, where there is an assumption of familiarity with the characters who recur on the global stage, and where the catastrophes that make up the various narrative arcs of current events take on an episodic quality. Outputs can be amusing in their lack of nous, or can seem almost satirical in their deadpan presentation of events that, when separated out from the character development for which the media has become the medium, come across as a little bit absurd. The system makes no effort to be topically selective, and reports on individualistic sports such as tennis and golf stripped of any mentions of the actual players are especially ridiculous.

But this sensation of the uncanny rediscovery of something known as something alien and unnamed, of being on the edge of a dream of a life that is completely unfamiliar, is what I set up The News to explore. In its current state, this project isn’t a very good way to keep on top of what’s going on in the world. But in some other version, could a move like this to devalue the currency of all these brands and personalities, and to give the world a chance to keep being new, be part of a cure for whatever it is that I think is making hard for some of us to get out of bed every day?

I can anticipate at least a few of what I suspect will be many reasonable objections that could be made to this project, one of which is this: isn’t one of the main points of journalism to hold individuals with power to account? To that, I can only respond that yes, this is how it should happen, but it’s really not working. Part of the problem is that simply saying, or showing, or proving that some great wrong has been done by someone in particular no longer really seems to have any meaningful consequence if that person has access to what were meant to be sources of truth and light. For all the evident enthusiasm in the parts of the world where the press is ostensibly free, there seems to be an epidemic confusion about the difference between the freedom of expression and systems of communication where actually getting heard is far from free.

Another objection is that the LLM-enabled summarisation process I’ve designed could be subject to inherent biases that get baked into the models. Maybe this happens by mistake, but one way or another, choices have to be made about how to reward models and how to tweak them, and anyway, like Howard Zinn said, “you can’t be neutral on a moving train”. In this respect an LLM is, at least in a relatively innocuous analogy, like an editor, who brings with them a voice, an overriding stylistic commitment, a sense of institutional mission, and along with all of this, inevitably, a world view. I don’t really know what to do about this. For now I’ve simply sought to prompt the summarisation model in a way that encourages nothing more than paraphrasing the content that is already in the source material for a story, but every minor stylistic decision can be construed as an ideology in disguise, and, when the author has no actual mind with which to hold beliefs or values, as is the case with LLMs, it becomes that much easier to imagine intent that can never be denied.

A third objection that some might raise to this kind of project is that it creates a barrier between human journalists and their human readers, to which I can only say… yes, that’s a point. For now, rather than really getting into this, I’ll offer this quote from Edward Herman and Noam Chomsky in their book Manufacturing Consent:

The elite domination of the media and marginalization of dissidents that results from the operation of these filters occurs so naturally that media news people, frequently operating with complete integrity and goodwill, are able to convince themselves that they choose and interpret the news “objectively” and on the basis of professional news values. Within the limits of the filter constraints they often are objective; the constraints are so powerful, and are built into the system in such a fundamental way, that alternative bases of news choices are hardly imaginable.

Some technical details

The News is regenerated every six hours, pulling stories from a set of established and ostensibly reputable news outlets. The sources for The News are:

Source	Type	Feed / API URL used
BBC News	RSS	http://feeds.bbci.co.uk/news/world/rss.xml
NPR News	RSS	https://feeds.npr.org/1001/rss.xml
Al Jazeera	RSS	https://www.aljazeera.com/xml/rss/all.xml
Ars Technica	RSS	https://feeds.arstechnica.com/arstechnica/index
Deutsche Welle	RSS	https://rss.dw.com/rdf/rss-en-all
France 24	RSS	https://www.france24.com/en/rss
CBC News	RSS	https://www.cbc.ca/cmlink/rss-topstories
ABC Australia	RSS	https://www.abc.net.au/news/feed/51120/rss.xml
The Guardian	API	https://content.guardianapis.com/search

As you can see, English-language RSS feeds are used in all cases except for The Guardian, which has a dedicated API specifically to facilitate automatic content processing. The scraping agent I’ve built tries to behave itself, using a legitimate user-agent to identify itself and respecting each feed’s terms of service and any available robots.txt.

This is the process that generates the stories you can see on The News:

1. Every six hours, all stories available for the last six hours from each of the above sites are pulled.
2. Top stories are selected using a round-robin system, grabbing one story from each of the sites outputs until 100 stories in total have been selected. In principle the round-robin is used to make sure there is a good distribution of stories between sources, and under this condition we count on stories returning whatever they consider to be their top story on top of the stack. In practice there are often less than 100 stories total.
3. An LLM is used to cluster the top 100 stories, grouping together any stories that are deemed to be about the same basic thing. A simple, generic description for each story-cluster is generated, which will become the headline for that story on The News.
4. The top 10 story-clusters are selected, ranked based on the total number of scraped news items contributing to each cluster.
5. An LLM is used to summarise each of the top 10 story-clusters into a single, coherent news item. The LLM is prompted to write these stories in a generic way, without bias and minimising reference to the names of individuals and corporations. The names of countries and, where required, institutions is not actively discouraged.
6. The News is updated with the 10 rewritten stoies.

I’m currently using the open-weight gemma-4-31b-it:free model LLM, developed by Google, for semantic intelligence, so this could in principle be instantiated locally with some fairly accessible hardware. In practice I’m running this through the OpenRouter API.