How I Built a Personal News Tamer to Cut Through the Noise

The Morning Scroll That Went Nowhere

Every morning, I do the same thing: grab my phone and see what happened overnight. WeChat has dozens of unread messages, Weibo trends have rotated a few times, news apps push a long list of “important updates,” and tech forums buzz with “just released” and “major update” posts.

After fifteen minutes of scrolling, a weird feeling creeps in: I know a lot of stuff, but I can’t name one thing that actually matters today. A phone launches, an AI model updates, a company announces a new plan. Each one looks worth tapping, but when I read them all, little sticks.

We used to worry about not having enough information. Now the problem is too much—so much that it feels like being swallowed by an expanding black hole, eating not just time but my judgment about what deserves attention.

The Experiment: Build a Personal News Filter

So I decided to run an experiment: use an AI agent to build a custom news filtering system, and take back control over “what to read each day.”

The first version was simple: a scheduled task at 8 a.m. that gathers five tech news items and sends them to me. It worked okay—instead of hopping between news sites, apps, and social feeds, I’d have a tidy summary waiting while my coffee brewed.

But the novelty wore off after a few days. The same model launch, written three different ways, would eat up three of my five slots. Yesterday’s news, retold by another outlet, would pop back as “latest.”

The agent wasn’t lazy. It was fed a bag of loose “bulk information”: official announcements, media reports, secondhand takes, reposts, clickbait—all looking equally fresh. Hand that to any AI, and it’ll summarize fast but can’t judge what’s genuinely worth your time.

Step One: Ditch the Firehose, Build a Tiered Source Pool

To cut the echo, I started with the sources. Over the years, I’ve collected 161 RSS feeds. But quantity isn’t quality. In an age of recycled content and AI-generated fluff, the core principle is “traceability.” Information loses critical context—and often its original meaning—when it’s endlessly repackaged. You want to get as close to the origin as possible.

So I sorted my feeds by trustworthiness:

• Primary sources: official blogs from OpenAI, Google DeepMind, Anthropic—raw, unaltered announcements.
• Serious media: Bloomberg, The Information, Business Insider, WSJ, Reuters—solid reporting with cross-checking.
• Quality secondary outlets: The Verge, Techmeme, TechCrunch, MacRumors—they turn dry technical news into readable stories.
• Bloggers and KOLs: tech reviewers on Weibo or Bilibili—for hands-on impressions and unique takes.

(If you’re into apps and gadgets, you can also subscribe to ifanr and APPSO—they’re great.)

Step Two: Organize the Stream with a Tree Structure

Managing over a hundred feeds gets messy if they’re all in one list. I use Folo, an RSS reader that lets me organize subscriptions into folders. It feels like flipping through a custom magazine—my feeds are divided into six sections: tech, gaming, culture, AI, autos, and more.

RSS might sound old-school in the age of algorithmic feeds, but it’s still the best way to actively aggregate and control what you see.

Step Three: Give Your Agent an “Outer Brain”

Folo also offers a CLI, so my agent can read unread items from my curated feeds—not random web grabs. Each item comes with a direct link, which eliminates the hallucinated URLs that LLMs sometimes invent.

My upgraded morning briefing now uses the Folo API to read unread items from the past 24 hours, then applies a few “editor-in-chief” rules:

• Merge duplicates: if multiple sources cover the same story, output it only once.
• Cross-verify: compare details from different sources and combine them into a richer summary.
• Prefer primary sources: if an official announcement exists, use its wording and note “multi-source verified.”

Step Two (Again): Teach Your Agent Your Taste

A good source pool isn’t enough. Industry hot topics aren’t necessarily your personal hot topics. For a while, open-source AI models dominated the news. Day one, I clicked. Day two, more parameter breakdowns. By day three, I knew it wouldn’t change my day—but the agent didn’t. It only saw that the topic was still trending.

So I told it directly: “Too much AI today. I want more consumer electronics and hardware news. 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 full of hardware news, no AI models in sight. That’s the part I love: the agent doesn’t magically understand you, but it can remember what you don’t like and what you care about. A good assistant is often “trained” by your complaints.

Step Three: Turn Your Feed into a “Cyber Newspaper”

Dialog boxes are fine for quick summaries, but reading is better with a bit of design. Since the agent can code, I asked it to turn the final five briefs into a clean HTML page—a minimalist card layout with a title, key facts (multi-source), why it matters, and source links as buttons.

The result: no more cramped text blocks. White cards present verified facts and commentary, and I can jump to the original source with one click.

I even used the same approach for a long-running topic: the foldable iPhone. Rumors, denials, more rumors—each one looks like a big deal, but together they’re just the same question rephrased. So I had the agent build a self-contained HTML “dynamic encyclopedia” that tracks the foldable iPhone rumors.

The page includes a 100-word status summary, a tree diagram of specs (screen ratio, hinge, price, etc.), a timeline of key leaks, a keyword frequency chart based on independent sources (not reposts), and rumor cards tagged as “confirmed,” “multi-source corroborated,” “single rumor,” or “unverified.” Each card links to the original source—like Ming-Chi Kuo or Bloomberg.

This cross-referencing saved me hours of hunting through pages. It handed me a structured report instead of a pile of tabs.

Why This Matters More Than Ever

After seeing that foldable iPhone tracker, I’m actually less anxious about when it’ll launch. Not because I’ve lost interest—no rumor page can replace holding the real thing. But when you can see which claims are backed by supply chain consensus and which are just clickbait, the fear of missing some “big news” fades.

That anxiety disappears when you cut out low-quality info. In 2025, Merriam-Webster chose “slop” as the word of the year—originally meaning “swill,” now used for AI-generated, low-quality digital content. More info, better packaging, harder to judge.

In this age of digital slop, building your own reliable sources is the best defense against AI garbage. AI can collect, dedupe, and organize, but it shouldn’t outsource your judgment.

The flood of information won’t stop. Instead of trying to swim faster, build a small dam: subscribe to sources you trust, keep diverse voices, and go back to the original when you see a claim. Whether it’s curating feeds or letting an agent follow your preferences, all we’re doing is putting a filter upstream.

What reaches you is already settled. You’ll naturally know what deserves a slow read and what’s just a glance. Extracting what’s genuinely useful from the noise—that’s a hobby worth keeping.