Podego: Using AI Responsibly to Inform and Empower
Before diving into how it works, let’s first explore what Podego does. Podego creates custom audio news episodes based on your personal preferences using AI tools. Think of it as your personal news aggregator minus the information overload, combining news from multiple sources into one simple objective story and delivering that story daily (or weekly) to your phone—narrated in a voice that resonates with you.
So, how does it work? 🤔
The Podego process consists of two key stages.
Part One: Finding the Story 🔎
Podego continuously monitors RSS feeds through specialist services, alerting us whenever a new article appears on a relevant topic. This could be a news article, a blog post, a podcast, or even a video—all sourced from trusted, diverse outlets.
Our data-gathering algorithm then reads and processes the content. Next, the system runs the article through Large Language Models (LLMs) and employs Natural Language Processing (NLP) techniques to extract a summary of key knowledge claims. This summary is rewritten as a concise script, containing only verifiable facts.
By rewriting stories into fact-based summaries, our system determines what a story is about and what claims it makes. To ensure accuracy, we apply the same knowledge-claim extraction process to our own Podego-generated summaries—preventing any accidental hallucination of new facts. This process forms the foundation of our knowledge extraction, leaving behind only the verified facts while filtering out opinions and falsehoods.
Next, we use a model designed to embed sentences and paragraphs into a high-dimensional vector space. Yes, it’s as futuristic as it sounds! Without giving away the special ingredient of our secret sauce, this method allows us to identify relationships between stories in our database. Often, you’ll find multiple versions of the same story in this database, told from different perspectives; meaning our stories don’t come from just one place.
By clustering related articles, we can combine them to present you with a more complete and less biased picture of events, gathered from multiple sources. What’s especially cool about this vector-based approach is that non-factual information—anything that doesn’t align in vector space—gets discarded automatically.
Part Two: Delivering the Story to the User 🎧
Once the stories are processed, a script is generated for each one. When a user requests an episode—either by clicking on a story or scheduling one for later —the system springs into action.
Using a recommendation algorithm that ranks articles according to a personalized scoring system, Podego matches users with the most relevant stories based on their preferences. The system then compiles the content into a tailored episode. 📝
Once the episode's content is finalized, LLMs generate the introductions and outros, and AI voices bring the script to life—on demand, in just a split second. 🤖
Since the story pool is constantly updating, the Podego app ensures that users receive the most up-to-date and accurate news possible. The result? A fully customized news experience delivered in the voice and style that suit you best, ready to play, every day.
More Than Just an Aggregator
Traditional RSS aggregators can be overwhelming. Podego is different—it filters out the noise, delivering a clean, curated stream of information that truly matters to you. It strikes the perfect balance between keeping you informed and avoiding information overload.
When it comes to AI in commercial applications, we believe in an ethical approach. We aim to deploy LLMs in service of people’s needs. Podego doesn’t seek to exploit —it amplifies the work of journalists and news organisations, making information more accessible and digestible for users.
If you’re interested, sign up for the Beta testing of the Podego app or subscribe to our Substack mailing list for updates and announcements. 📬
