How I Built an AI News Editor That Filters 161 RSS Feeds

The Problem: Too Much News, Too Little Signal

Every morning, I used to wake up and do the same thing: grab my phone and see what happened overnight. Dozens of unread messages in WeChat. Twitter trends already churning. News apps pushing a stream of “important” alerts. Then I'd hop between tech forums and catch the latest “just announced” or “major update.”

After ten minutes of scrolling, a strange feeling settled in. I knew a lot of things, but I couldn't tell you what actually mattered today.

A phone launches. An AI model updates. A company announces a new plan. Each item looks worth clicking, but once you've clicked them all, little sticks. We used to worry about not having enough information. Now the problem is the opposite: there's so much that it swallows not just your time, but your ability to judge what's worth your attention.

So I ran an experiment: build a personal news filtering system with an AI agent. Take back control of what I read each day.

Step One: Stop Scraping the Whole Web, Start with Trusted Sources

My first attempt was simple: set up a scheduled task in my agent to collect five tech news stories every morning at 8 AM. That gave me a basic briefing assistant.

It worked, sort of. Instead of jumping between news sites, apps, and microblogs to piece together what happened in tech overnight, I could make coffee and find a summary waiting on my phone.

But the novelty wore off in days. The same product launch, written three different ways, would occupy three slots. A story I'd already read yesterday would resurface as “latest” because another outlet rewrote it.

I couldn't blame the agent. When it searches the whole web, it gets a grab bag of raw info: press releases, news articles, secondhand commentary, reposts, clickbait—all looking equally “new.” An AI can summarize that fast, but it can't judge what's genuinely important.

To cut through the echo chamber, I started with the sources themselves. Over the years, I'd collected 161 RSS feeds. But quantity isn't quality. To stay sane in this information flood, I needed a clear filtering system.

Action 1: Build a Tiered Source Pool by Credibility

The core principle is “source tracing.” When information gets passed around and aggregated, it loses context and can drift from the truth. The closer you get to the original node, the clearer the picture.

So I organized my feeds by trust level:

• Primary sources – official company blogs like OpenAI News, Google DeepMind News, Anthropic News. These give you the raw, unedited announcement.
• Serious media – Bloomberg, The Information, Business Insider, WSJ, Reuters, Caixin. They have real reporting and cross-checking.
• Secondary aggregators – The Verge, Techmeme, TechCrunch, MacRumors. They turn dense info into readable stories.
• Bloggers and KOLs – Weibo tech influencers, Bilibili reviewers. Good for hands-on impressions and unique takes.

(If you don't mind the plug, subscribe to ifanr and APPSO too—they're solid.)

Action 2: Organize Your Feeds Like a Magazine

Once you have hundreds of feeds, you need a container. I use Folo, an RSS reader that lets you arrange subscriptions into categories. It feels old-school in the age of algorithmic feeds, but for actively pulling and controlling information, nothing beats it.

In Folo, I've split my sources into six sections: tech, gaming, culture, AI, autos, and so on. Opening it feels like flipping through a magazine I've personally curated.

Action 3: Turn Folo into an Agent's Brain

The real game-changer is Folo's CLI. It turns my 161 feeds into a database the agent can query directly. Instead of scraping random headlines, the agent reads my curated unread items. Each item even includes a direct link, so there's no risk of the AI inventing URLs.

Here's the upgraded morning briefing prompt:

Use Folo. Read https://api.folo.is/skill.md and follow the instructions. Read my unread items from the past 24 hours. Apply “editor-in-chief” judgment: merge duplicate stories, cross-check details, prioritize official sources, and mark multi-source verification.

Step Two: Teach Your AI What You Actually Care About

Even with a good source pool, industry hot topics aren't the same as your personal interests. For a while, open-source AI news dominated my feed. Day one, I clicked. Day two, I read the parameter breakdown. Day three, I knew it wouldn't change my day.

What I really cared about were concrete hardware moves: a new laptop's specs, a phone's release date. But the agent kept pushing AI news because it was hot.

So I told it directly: “Too much AI today. I want more consumer electronics and hardware. Remember this for future briefings.”

The agent created a MEMORY.md file and updated its long-term rules. And it worked—the next briefing was all consumer tech.

That's what I love about agents: they don't read your mind, but they remember what you don't like. A good assistant is often trained by getting scolded.

Step Three: Turn a Chat Feed into a Digital Newspaper

Dialog boxes are fine for quick summaries, but reading a wall of text gets tiring. Since AI can write code, why not have it generate a custom HTML page?

I upgraded the prompt:

Take the top 5 deep-dive summaries and render them as a clean, card-based HTML page. Include: title, core facts (multi-source), why it matters, and source buttons.

The result is a clean white-card layout with buttons linking to original sources. No more compressed text. It feels like a newspaper designed just for me.

I pushed the idea further with a long-running topic: the foldable iPhone. For months, rumors, denials, and re-rumors looped endlessly. Each headline felt huge, but together they were just the same question rephrased.

So I had the agent build a “living encyclopedia” for the foldable iPhone:

Create a self-contained HTML page tracking foldable iPhone rumors. Include: a 100-word status summary, a tree diagram of specs (size, hinge, price), a timeline of key leaks, a keyword frequency chart based on independent sources, clue cards with credibility and status, filtering by category/credibility/status, and a source list. Use official Apple, supply chain, analyst reports, and original media. Don't count reposts as independent evidence. Distinguish confirmed, multi-source corroborated, single rumor, unverifiable. Output only HTML with inline CSS/JS, dark theme, responsive.

The agent produced a structured dossier. At the top, a 100-word summary with no fluff—just where mass production stands and what's still unknown. Below, a tree diagram mapping every rumored spec, from screen ratio to liquid metal hinge to price.

More valuable than the specs is how it weaves fragments together. The timeline and keyword frequency chart show how a rumor evolves and where consensus forms. Each clue card is tagged “Confirmed,” “Multi-source,” or “Single Rumor.” For something like “no crease,” I can see how many independent sources back it and click through to Ming-Chi Kuo or Bloomberg directly.

This cross-referencing saves me from piecing together the truth from a hundred tabs. It hands me a clean investigation report.

After seeing that page, I'm actually less anxious about when the foldable iPhone will launch. Not because I've lost interest—rumors can't replace holding the real thing. But when you can see which claims are supply-chain consensus and which are just clickbait, the fear of missing out fades.

Why This Matters: Fighting the Slop

In 2025, Merriam-Webster chose “slop” as its word of the year. It used to mean pig feed, but now it describes the flood of low-quality AI-generated content.

There's more information than ever, and it's better packaged, but judgment gets harder. Building a reliable source network is the best defense against AI slop.

The point isn't to outsource your judgment to an AI. It's to use AI to collect, deduplicate, and organize—while you keep the final say.

The information flood isn't going away. Instead of trying to swim faster, build a small dam upstream: subscribe to sources you trust, keep diverse voices, and always check the original when you see a conclusion.

Whether it's handpicking feeds or teaching an agent your preferences, what you're really doing is installing a filter at the top of your information stream. What reaches you is already sifted. What deserves a deep read, and what deserves a quick skim, becomes obvious.

Being able to pull what's useful out of the noise—that's a skill worth keeping.