Tech News
← Back to articles

Particle’s AI news app listens to podcasts for interesting clips so you you don’t have to

read original related products more articles

An AI news app called Particle, from former Twitter engineers, can now keep up with news breaking on podcasts as well as news published on the web.

Just ahead of its recent Android release, Particle has introduced a feature called Podcast Clips, which finds the most interesting and relevant moments across many different types of podcasts, and then includes those clips alongside the related news stories in its feed.

So instead of listening to a lengthy podcast just to catch the 45 seconds of interesting comments, you can play back the clip as you’re reading the news on Particle. You also have the option of reading the transcript of the clip instead, as the words are highlighted as they’re spoken.

Image Credits:Particle

“We’ve done that basically for any news story — if there is a podcast that is talking about it, or relevant at all, we’ve got all those clips,” Particle CEO Sara Beykpour, previously the Senior Director of Product Management at Twitter, told TechCrunch. “It’s a really cool way, when you’re reading a story or learning about a story, to get a breath of what are people saying about this? What’s the commentary?”

The addition acknowledges a shift in the news ecosystem that’s been underway for years. Not only are more people getting their news from podcasts and trusting them as reliable sources, but the medium is also becoming a destination for breaking news and major announcements from public figures.

Tech CEOs, in particular, are now seeking out friendly podcast hosts to air their talking points instead of trying to work with traditional media, as Bloomberg reported in 2024.

That makes paying attention to podcasts even more critical if you want to keep up with news.

Beykpour says Particle uses embedding models to understand when podcasts relate to a given news story. These models are provided by the same companies that provide LLM models, but they’re not generative AI technologies, she explains.

“We use vector embeddings to understand that these different parts of the podcasts are related to these different stories,” Beykpour notes. “A single podcast might cover 10 or 20 stories, so we use AI to understand that. We also use AI to do some of the logic around clipping, and understanding when to start a clip and end a clip.”

... continue reading