Particle App Uses AI to Clip Podcast Highlights
Particle, an AI-powered news app from former Twitter engineers, now extracts relevant short clips from podcasts and pairs them with related news stories. The app uses vector embedding models to match content and offers premium features like custom summaries for a subscription fee.
- The app was co-founded by Sara Beykpour, former Senior Director of Product at Twitter, and Marcel Molina, a former senior engineer at both Twitter and Tesla. - Particle has raised a total of $15.3 million, consisting of a $4.4 million seed round and a $10.9 million Series A led by Lightspeed Venture Partners. - The technology stack includes a hybrid of AI models for its features, such as OpenAI's GPT-4, Anthropic, and Cohere, while transcription for the podcast clips is handled by ElevenLabs. - To ensure factual accuracy for its core content matching, the app uses vector embedding models for semantic analysis rather than generative AI, which the company claims reduces the likelihood of AI errors from one percent to one in ten thousand. - The premium subscription, Particle+, costs $2.99 per month or $29.99 per year. - Premium features include the ability to request summaries in different styles like "Explain Like I'm 5," choose from multiple voices for the audio feed, and ask the in-app AI private questions about a news story. - The company has established content partnerships with several major news organizations, including Reuters, Agence France-Presse (AFP), and Fortune, to integrate their content via APIs. - In addition to iOS, the app recently launched on Android, expanding its potential user base.
