Particle's Radar Turns 130,000 Podcasts Into Searchable, Queryable Text
Particle, the AI newsreader startup founded by former Twitter engineers, launched Radar on August 26, 2026. Radar is a podcast search engine that transcribes podcast audio into text, uses AI to understand the meaning of what's being said, and pulls out key quotes and highlights. It is hosted at radar.particle.pro. TechCrunch
The service transcribes more than 130,000 podcasts, which Particle describes as the largest transcribed podcast index in existence. Its catalog includes all of Apple's Top 200 podcasts across 135 categories, and it adds 20,000 new episodes daily. Transcriptions include speaker labels (identifying who is talking) and detailed metadata, and the system performs entity recognition — automatically identifying people, companies, brands, products, and topics mentioned in each episode. TechCrunch
Radar's origins are internal. The feature grew out of a capability in Particle's news-reading app that surfaced interesting podcast clips. The company had been analyzing thousands of episodes per day to identify top clips as far back as May 2026, when it described the work in a public post. Radar turns that pipeline into a standalone product with a searchable index and an API — a programming interface that lets other software connect to Radar's data directly. TechCrunch; Threads
Beyond search, Radar can track mentions of specific people, companies, or topics across podcasts and send alerts when they appear. Alerts can be configured for real-time delivery at mention time or as daily or weekly digests, and they can be delivered via email, Slack, or webhook (an automated message sent from one application to another). Users can filter alerts by guest, topic, or a restriction to top-tier podcasts. Radar also extracts self-contained clips with timestamps, so users can both listen to and read the relevant passage. TechCrunch
The customer base reveals where this kind of structured podcast data has the most immediate pull. Hedge funds are the highest-volume customers directly integrating with Radar's API. Other top-paying customers include AI search platforms and data resellers. Exa, a search API provider built for AI agents, is among Radar's integration partners. TechCrunch
Particle also offers a Podcast Intelligence API that lets users search for podcast mentions across episodes and shows, updated daily. That API, alongside the Radar web interface, forms the two surfaces through which the underlying transcription and entity-extraction pipeline reaches external consumers. Threads
The broader context here is the shift from podcasts as a listen-only medium to podcasts as a queryable text corpus. Podcasts have been a notoriously difficult content type for automated systems to work with: audio is unstructured, episodes can run for hours, and the valuable moments are buried inside long-form conversation. Radar's transcription-plus-entity-recognition approach turns that audio into structured records that machine systems can search, filter, and act on.
That hedge funds lead in API volume is not surprising. Earnings calls and analyst interviews have long been mined for trading advantage, but podcasts increasingly host the same kind of forward-looking commentary from executives, investors, and industry operators. A fund that can receive a Slack alert the moment a portfolio company's CEO names a competitor on a top-200 show has a material information advantage over one relying on next-day transcripts.
The fit with AI search platforms and data resellers is equally straightforward. AI agent systems need structured, up-to-date sources to reason over, and podcast content, with its conversational tone and frequently unscripted disclosures, fills a gap that text-only web scraping cannot. A webhook-delivered entity mention with a timestamped clip is exactly the shape of input that a retrieval-augmented pipeline — a system that pulls in external data to give an AI model relevant context — can consume without further preprocessing.
One question worth flagging is the competitive moat. Transcription is commoditizing, and automated speech recognition quality continues to improve across providers. Radar's defensibility, if it has one, lies less in transcription accuracy than in the scale of its index and the entity layer built on top of it. 130,000 shows and 20,000 daily episodes is a substantial ingestion operation, and the metadata, speaker labels, and entity tracking add processing value that raw transcription does not. Whether that is enough to hold off larger search or audio platforms that decide to build the same thing is an open question.
Particle's CEO and co-founder is Sara Beykpour. The company's trajectory, from a news-reading app with a podcast-clip feature to an API business serving quantitative funds and AI search infrastructure, suggests a deliberate pivot toward the infrastructure layer — the behind-the-scenes plumbing that other businesses build on. That is where the revenue is concentrating, and the product roadmap, with its webhook integrations and agent-friendly API, is built accordingly. TechCrunch
For now, Radar is the only service operating at this transcription scale with an API designed for programmatic consumption by trading desks and agent-based systems. Whether it stays that way depends on how quickly the larger audio and search platforms decide that structured podcast intelligence is worth building in-house.
