Not the 9 O'Clock News. Building an aggregator you can trust
There’s no shortage of ways to consume news in 2025. Apple News, Google Discover, Flipboard, Twitter/X feeds, algorithmic TikTok recaps—every service promises to surface what matters. And yet, more often than not, what I end up scrolling through is a mess: half-baked clickbait, PR puff pieces, and breaking-news alerts that I don’t actually need to break my morning coffee over.
So I decided to build something different. Not a competitor to the big platforms, but a personal news system. Something that could filter stories on a granular level, pull out only what I actually want to know about, and then wrap it all in a bit of editorial framing so that I don’t just get the raw feed—I get perspective.
The result isn’t a glossy app ready to be downloaded from the App Store. It’s a working flow. A chain of logic and automation that delivers an email to me every morning with three or four pieces of news, summarised and put into context in a way that I’ll actually pay attention to. No fluff. No doomscrolling. Just the information I care about, in a voice I recognise.
This post is about how that works, why it matters, and why I think more of us should stop relying on “default feeds” and start building our own.
Apple News is decent. Google Discover is fine. Twitter/X is chaotic but sometimes useful. But all of them share the same problem: they are built for the average.
The “average” Apple News reader wants a blend of politics, lifestyle, sport, and celebrity gossip. The “average” Google Discover user wants trending topics, a smattering of local news, and a dash of whatever tech giant has paid for visibility that week.
That’s not me.
I want to know about AI policy shifts that might affect my work, creative technology experiments that inspire me, and the occasional big-picture geopolitical move. I don’t need celebrity diets, football transfer rumours, or half-page articles about which biscuit best matches your star sign.
So what happens? I scroll, skip, scroll, and skip again. Eventually, I hit something useful, but I’ve already wasted time and attention.
The issue isn’t that the content isn’t out there. It’s that I can’t filter finely enough. I want to tell the system: “Show me AI regulation, but only where it intersects with Europe or the UK. Show me stories on built-environment technology, but only when they go beyond press-release hype. And give me one or two things I wouldn’t normally click on, but explain why they’re worth a look.”
Mainstream platforms aren’t designed for that. They’re designed for broad engagement. Which is why building your own flow is so powerful.
The backbone of my system is simple: a set of nodes (I’m using n8n, but you could do this in Zapier, Make, or even custom code) that pull in news from different feeds, filter it, transform it, and then package it into something readable.
Here’s the rough shape:
Trigger
Every morning at 7am UK time, the flow starts. This is my personal “front page” moment. Not whenever I open Twitter. Not whenever an app decides to ping me. My time, my terms.
Fetch Articles
I pull in feeds from trusted sources: RSS feeds from tech policy sites, AI newsletters, security industry updates, and a couple of general outlets for context. The key is variety but within a bounded domain.
Filter & Deduplicate
This is the bit the mainstream apps can’t do well. I apply filters like:
Must include keywords around AI, automation, or security.
Exclude “press release” style phrases (“announces new partnership”, “brand launch”, etc.).
Remove duplicates, so if two outlets are reporting the same wire story, I only get one.
Style Picker / Editorial Layer
This is where the LLM comes in. Instead of just dumping links, I feed the article snippets into a model and ask it to:
Summarise the story in plain English.
Highlight why it might matter (context, implications, questions).
Offer me two outputs: one more formal for LinkedIn, one punchier for Twitter/X.
This gives me a bit of editorial perspective, in my own “house style,” without me having to spend an hour rewriting everything.
Compose & Email
Finally, everything gets packaged into a simple HTML email:
Title, source, and link.
A snippet or summary.
Draft posts ready for social, so I can copy-paste if I want to share.
I wake up, check my inbox, and there it is: my curated digest, written for me, by me.
The filtering alone is useful. But the real magic is in the editorial layer.
Because here’s the thing: raw feeds give you stories, but not perspective. And perspective is what helps you decide whether to care.
Take a headline like:
“UK positions itself as global AI superpower with new investment package.”
On its own, that’s noise. Another puffed-up press release, easily skipped.
But add a layer of editorial:
“Billions are being poured into AI, but the question isn’t whether the UK will become a superpower—it’s whether small businesses will actually see any benefit. If you’re in a smaller organisation, the smart play isn’t chasing hype, it’s figuring out how to integrate AI tools practically and cheaply.”
Now the story has weight. It’s not just news, it’s relevance.
That’s why I ask the system not just to summarise, but to contextualise. To add a “why it matters” line, in my voice. And because the model has been fed enough examples of my past writing, the end result feels familiar. Like I’ve already half-written it.
There’s a bigger point here beyond just the tech. It’s about building your own filters instead of just consuming what’s given.
When you use Apple News, you’re downstream of someone else’s editorial logic. When you build your own aggregator, you move upstream. You define what makes it through.
This shift matters because attention is scarce. Every headline you read, every scroll you make, is a tax on your focus. By filtering before the stories ever hit your eyes, you reclaim that focus.
And once you’ve built the pipeline, you can tweak it endlessly. Want more global politics? Add another feed. Too much fluff? Tighten the filters. Need a change of voice? Adjust the editorial prompts.
The act of building gives you control, and the control makes the output more valuable.
Apple News and its peers aren’t going anywhere. They’re good for broad strokes. If you want to know what the mainstream is paying attention to, they’re useful.
But they’re not built for nuance. They’re not built for people who want to go deep into specific niches or filter based on subtle criteria.
Think of it like cooking: Apple News is the supermarket meal deal. Cheap, quick, broadly fine. My system is cooking at home. Takes a bit of prep, but you decide the ingredients, the seasoning, the portion size.
And once you’ve tasted the difference, it’s hard to go back.
A few lessons from actually running this flow for the past few weeks:
Three is the magic number. More than three stories in the daily email, and I stop paying attention. Less than three, and it feels too thin.
Titles matter more than you think. I didn’t realise how much I skim using headlines until I saw “(no title)” appear in my email. Pulling through the original article title makes the whole thing feel anchored.
Context beats quantity. I’d rather have one well-explained story than ten bare links. The editorial layer isn’t just a nice-to-have, it’s the whole point.
Time zones are tricky. Running a 7am UK trigger when your workflow host thinks in New York time leads to some unintentional midday deliveries. Worth double-checking before you rely on it.
It feels less like “news” and more like “briefing.” Which is exactly what I want. Not a firehose, not a feed. A briefing, for me, by me.
Right now, the system ends with me: I get the digest, I sometimes share posts, and that’s it. But there’s plenty of scope to take it further:
Collaborative filtering: What if a small group of us pooled our flows, so we got not just our own filters but also recommendations from trusted peers?
Sentiment tagging: Adding a quick “positive / negative / neutral” label to stories could make it even easier to skim the mood of the day.
Voice experimentation: Different editorial voices for different contexts. Serious for LinkedIn, irreverent for personal notes, concise for Twitter.
Integration with note-taking apps: Automatically sending the most relevant snippets into Notion or Obsidian to build a long-term knowledge base.
This is just the start. The beauty of building your own pipeline is that you can keep iterating.
We’ve all become used to news as something that happens to us: notifications, alerts, endless scrolls. But it doesn’t have to be that way.
With a little bit of setup, you can flip the relationship. You can decide what counts as news. You can add your own editorial voice. You can reclaim attention from the noise.
For me, the result is a daily email that feels more like a conversation with myself than a broadcast from a faceless algorithm. It tells me not just what happened, but why I might care.
And in a world drowning in headlines, that’s the only kind of news worth having.
If you want help setting up your own aggregator please reach out to ikirugai!
