How I Built an AI News Editor to Tame the Earth Science Information Flood

Morning Dread in the Data Deluge

Every morning, I used to wake up to the same ritual: grab my phone, swipe through dozens of messages, scroll past a hundred headlines, and still feel like I'd missed something important. The feeling got worse when I switched from tech news to Earth science—because now the stakes felt higher. A new wildfire report, a seismic shift in policy, a climate model update—each one looked urgent, but after an hour of reading, I couldn't tell you what actually mattered.

The problem isn't scarcity anymore. It's the opposite. We're drowning in so-called updates, and the noise is getting louder. AI-generated summaries, press releases dressed as journalism, and recycled coverage all blur together. The signal gets buried under a pile of digital slop. I needed a way to cut through that noise, so I built a system that treats Earth science news like a breaking story—not a firehose.

Step One: Stop Scraping the Internet, Start Curating Sources

My first attempt was naive. I set up an AI agent to search the web every morning and send me five news items. It worked for a day. Then I realized the agent was just grabbing whatever was popular—same story, three different rewrites, all cluttering my inbox. The problem wasn't the AI; it was the input. I was feeding it a pile of unvetted links and expecting it to know what mattered.

So I went back to basics. I built a curated list of 161 RSS feeds, but this time I organized them by trust level. For Earth science, that meant putting primary sources first: USGS earthquake alerts, NOAA climate bulletins, NASA Earth Observatory updates. Then came serious outlets like Nature, Science, and Eos. After that, specialized blogs and analysts who add context. Finally, a few voices on social media for on-the-ground observations.

Building a Tiered Source Pool

The key is to treat sources like a pyramid. At the top, you have official data and peer-reviewed papers. Below that, reputable journalism that verifies claims. Then comes interpretation and commentary. Everything below the top tier is suspect until it cites something above it. I applied this hierarchy to every feed I subscribe to, and I cut out anything that didn't clearly trace back to a primary source.

For example, when a new earthquake hits, I want the USGS shakemap, not a blog that reposts it. When a climate report comes out, I want the original PDF, not a summary that might misquote it. This sounds obvious, but it's easy to let convenience override rigor. My rule now: if a story doesn't link to a primary source, I don't read it.

Step Two: Use an RSS Reader That Speaks to AI

Managing 161 feeds in a normal RSS reader is a nightmare. I needed something that could organize them into folders and also let my AI agent read them directly. I found a tool called Folo—it's a modern RSS reader with a command-line interface. That CLI was the game-changer.

Now my agent doesn't scrape the web blindly. It reads the unread items from my carefully curated feeds, which are already filtered by me. It also sees the source URL for each item, so it can verify the origin. No more hallucinated links. The agent knows exactly where each piece of information came from.

Step Three: Teach the AI What You Care About

Even with good sources, the AI didn't know my priorities. At first, it kept sending me stories about volcanic activity in Iceland when I cared more about hurricane season. That's when I realized I had to train it, not just configure it.

I started telling the agent what to focus on: 'more on sea-level rise, less on individual storm updates.' The agent wrote these preferences into a memory file, and it started adjusting its selections. It's like having a junior editor who learns your beat. You have to correct it a few times, but eventually it gets the pattern.

Step Four: Turn the Output into a Visual Briefing

Plain text lists are boring and hard to scan. So I asked the AI to build an HTML page for me every morning. It creates a clean card layout with a headline, a summary of key facts, and a 'why it matters' section. Each card has buttons linking to the original sources. This turned my morning briefing into something I actually look forward to opening.

For bigger stories, I use a similar approach to build what I call 'living trackers.' When a major story unfolds—like a slow-moving hurricane or a volcanic crisis—I have the agent create a dedicated page that tracks every new development. It includes a timeline, a map of reported facts, and a filter to show only confirmed info versus rumors.

Real-World Example: Tracking the Hunga Tonga Eruption

When Hunga Tonga–Hunga Ha'apai erupted in January 2022, I remember the chaos of trying to piece together what was happening. Reports conflicted, data was scattered, and misinformation spread fast. I wished I had this system then.

So I tested it on a hypothetical: I asked the AI to build a tracker for the 2022 eruption, using only primary sources like USGS, NOAA, and university seismology labs. The result was a structured page with a timeline of seismic events, a breakdown of confirmed facts versus unverified reports, and links to the original data. It wasn't perfect, but it showed me how much clearer the picture becomes when you filter out the noise.

Why This Matters for Earth Science

Earth science is uniquely vulnerable to misinformation. The data is complex, the stakes are high, and the public often relies on secondhand summaries. When a study drops, it's easy for a headline to misrepresent a small caveat as a major finding. By building a personal news system that prioritizes primary sources, I'm not just saving time—I'm protecting my understanding of the world.

This isn't about becoming an expert overnight. It's about having a reliable filter. The AI does the heavy lifting: it aggregates, deduplicates, and cross-references. But the judgment about what matters—what to read deeply, what to skim, what to ignore—that stays with me.

In a world where AI-generated slop is spreading, the best defense is a good offense. Take control of your information diet. Build a source list you trust, teach your tools to respect it, and always trace a claim back to its origin. The flood of data isn't going to stop, but you can build a dam.

The next time you see a headline about a melting glacier or a new fault line discovery, ask yourself: where did this come from? If the answer isn't a primary source, keep digging. Your understanding of the planet depends on it.