type
status
date
slug
summary
tags
category
icon
password
Using CrewAI, I created a virtual newsroom with a Researcher, Editor, and Writer to fetch, filter, and summarize AI news from trusted RSS feeds, with plans to enhance the output’s structure for better readability.
Last time, I shared how I used a large language model (LLM) to tame my chaotic email inbox (link here), spotlighting the messages that actually matter. That got me thinking—could I use LLMs to streamline another part of my life? Like keeping up with the flood of AI industry news without drowning in it. Here’s how I leveraged LLMs and an agentic workflow to create a personalized weekly AI digest, with a dash of simple and a sprinkle of structure.
The Big Idea
I consume a ton of AI-related content—blogs, newsletters, but not all news is created equal, and wading through AI-generated fluff on the internet? No thanks. I wanted a system to fetch the latest updates, filter out the noise, summarize the good stuff, and present it in a neat, readable format. Sounds like a job for an LLM-powered team, right?
Instead of cramming everything into one monster prompt (which, let’s be honest, would probably choke), I turned to an agentic workflow. Think of it like a newsroom: journalists gather the scoops, editors pick the best ones, and writers craft the story. I used the CrewAI package to orchestrate this, assigning roles to handle specific tasks. Here’s the breakdown:
- Researcher: Scours trusted sources for the latest AI news.
- Editor: Decides what’s worth my time (first).
- Writer: Summarizes the news into bite-sized, digestible chunks.
Tackling the Noise Problem
The internet’s bursting with AI-generated content, and I didn’t want my digest cluttered with it. So, I limited my sources to a handpicked list of trusted RSS feeds. This way, I’m only getting updates from outlets I know deliver the good stuff. It’s like curating my own little AI newsstand.
How It Works
Using CrewAI, I set up my virtual newsroom. The Researcher pulls articles from my RSS list, the Editor ranks them based on relevance (tuned to my interests, of course), and the Writer whips up concise summaries. I ran a quick test, and while it’s not perfect yet, the first version already delivers a solid digest with enough info to keep me in the loop. If something catches my eye, I can dive deeper on my own.
Room for Improvement
The output’s great, content-wise, but I’m picky about presentation. I want my digest to be structured—think clean sections, bullet points, maybe even a fancy table for readability. So, my next step is tweaking the Writer’s output format to make it pop. Picture this: a weekly email with a sleek, scannable list of AI breakthroughs, each with a quick summary and a link to the source. That’s the dream.
Why This Rocks
This setup saves me hours of scrolling and sifting. It’s like having a personal assistant who knows exactly what AI news I care about, without the coffee runs. Plus, it’s fully customizable—want to add a new RSS feed or tweak the “importance” criteria? Done.
So, there you have it—a fun, LLM-powered experiment to keep me ahead of the AI curve. I’ll keep tinkering to make it even slicker, but for now, I’m stoked to have my own AI digest machine.
