How I Built an AI News Editor with 161 RSS Feeds to Save My Mornings

Why I Needed a News Filter (and You Might Too)

Every morning, I used to wake up to a tsunami of notifications. Dozens of unread messages on WeChat, a dozen trending topics on Weibo, a long list of "important updates" from news apps, and then the tech forums with their "just announced" and "major release" headlines. Fifteen minutes of scrolling later, I felt oddly empty. I'd seen a lot, but couldn't tell you what actually mattered that day.

We used to worry about not having enough information. Now the problem is the opposite: there's too much, and it's swallowing our time and our judgment. So I decided to build my own AI-powered news curation system to take back control of what I read each day.

Step 1: From Random Scraping to a Quality-First Feed

My first attempt was simple: an AI agent that ran at 8 a.m., collected five tech news items, and sent them to me. It worked, sort of. Instead of jumping between websites and apps to piece together what happened overnight, I had a briefing waiting while my coffee brewed. But the novelty wore off in days. The same product launch, written three different ways, would hog three slots. Yesterday's news resurfaced as "latest" just because another site rehashed it.

The problem wasn't the agent's effort. It was the raw material. A web search gives you a bag of "loose information": press releases, media reports, secondhand takes, reposts, clickbait—all looking equally "new." To cut through the echo, I went back to basics: I curated a list of 161 RSS feeds I've collected over the years. But quantity isn't quality. I needed a system to manage them.

Building a Trust Ladder

My core principle: go to the source. Information loses context and accuracy with every retelling. So I ranked my feeds by trustworthiness:

• Primary sources: official company blogs like OpenAI News, Google DeepMind, Anthropic News—for unvarnished updates.
• Established media: Bloomberg, The Information, WSJ, Reuters, Caixin—for rigorous reporting and cross-verified angles.
• Quality secondary sources: The Verge, Techmeme, TechCrunch, MacRumors—they turn raw info into readable depth.
• Bloggers and KOLs: tech influencers and YouTubers—for hands-on impressions and unique takes.

If you want, you can add sites like ifanr and APPSO—they're solid too.

Organizing the Chaos with a Tree Structure

Managing over a hundred feeds is a chore. I use Folo, an RSS reader that lets me organize subscriptions into folders. It's old-school, but it's still the best way to actively aggregate information without letting an algorithm decide for me. My Folo is split into six main sections—tech, gaming, culture, AI, autos—so opening it feels like flipping through a magazine I've curated myself.

Making the Agent Talk to Folo

Folo also has a CLI, which means my AI agent can read my unread items directly. Instead of feeding it random web scraps, it gets a pre-filtered list of sources. Each item includes a direct link, so the AI can't hallucinate URLs. That was a game-changer.

Step 2: Training My Assistant to Know What I Like

Even with good sources, the agent didn't know my personal interests. For a while, open-source model news dominated my briefing. Day one, I clicked. Day two, more parameter breakdowns. By day three, I knew it wasn't relevant to my life. But the agent kept sending it because the topic was still hot.

So I told it directly: "Too much AI news today. I'd rather see consumer electronics and hardware. Remember that for future briefings." The agent wrote a MEMORY.md file and updated its long-term rules. From then on, it filtered accordingly. It actually worked—the next briefing was full of hardware news instead.

That's the beauty of an AI assistant: it doesn't read your mind, but it remembers what you don't like. A good assistant is often trained by your complaints.

Step 3: Turning the Briefing into a Cyber Newspaper

Text in a chat window is fine, but it's not pretty. Since the AI can code, I asked it to turn the five daily briefs into a clean HTML page—cards, minimal design, with sections for headline, key facts (multi-source verified), why it matters, and buttons linking to the original sources. Now I get a beautiful, scannable page every morning.

Going Deeper: Tracking a Long-Running Story

I also used the same system to track a long-running saga: the foldable iPhone. Rumors, denials, more rumors—each one looks like a big deal, but they're often just variations on the same theme. So I challenged the agent to build a "dynamic encyclopedia" page for it.

The result was a self-contained HTML file with a 100-word status summary, a tree diagram of specs (from screen ratio to hinge material to price), a timeline showing when each leak first appeared and how it changed, a keyword frequency chart counting independent sources (not reposts), and filterable cards for each claim, labeled with confidence levels like "confirmed" or "single rumor." Everything linked back to the original source, whether it was Ming-Chi Kuo or Bloomberg.

That page didn't make me more impatient for the launch. In fact, it calmed me down. When you can see which rumors have solid backing and which are just clickbait, the fear of missing out evaporates.

Why This Matters Now More Than Ever

In 2025, Merriam-Webster chose "slop" as its word of the year—referring to low-quality, AI-generated content flooding the internet. The more information there is, the harder it is to judge.

Building your own reliable source list is the best defense against AI slop. The AI can collect, deduplicate, and organize, but it can't outsource your judgment. Instead of trying to swim faster in the flood, 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 curating feeds or teaching an agent your preferences, what you're really doing is installing a filter at the top of your information stream. What flows down to you is already settled. You can tell at a glance what's worth reading deeply and what's just noise. And that ability—to extract something useful from the noise—is worth more than ever.