Particle, the AI newsreader startup built by former Twitter engineers, just pivoted to something with bigger money behind it: making the spoken word inside podcasts searchable. According to TechCrunch AI, the company launched Radar on Wednesday, a podcast search engine that transcribes audio and actually understands what’s being said, so it can surface key quotes, clips, and highlights. What stands out here is who’s already paying for it. TechCrunch AI reports that hedge funds have been the highest-volume customers integrating directly with the API, hunting for data their trading agents simply can’t reach.
That last point is the real story. Most tools that feed AI agents crawl the open web, and the web is text. Audio is a blind spot. “Agents are generally blind to audio; they can’t see it unless something or someone has transcribed it,” Particle co-founder and CEO Sara Beykpour told TechCrunch. Radar is a bet that whoever indexes spoken media first owns a layer nobody else has.
Here’s what Radar actually does, based on the TechCrunch AI report:
- Transcribes at massive scale. Radar covers more than 130,000 podcasts, which the company calls the largest transcribed podcast index in existence. That includes every show in the Apple Top 200 across 135 verticals, with roughly 20,000 new episodes added daily.
- Understands entities, not just words. Transcriptions come with speaker labels and rich metadata. Radar recognizes the people, companies, brands, products, and topics being discussed, so you can search meaning instead of guessing at keywords.
- Tracks mentions and sends alerts. You can follow a person, company, or topic and get pinged the moment it comes up, or as a daily or weekly digest. Alerts land via email, Slack, or webhook, and filters let you narrow things down, like only flagging when a specific guest appears and talks about a specific subject.
- Pulls self-contained clips with timestamps. Radar pre-selects notable moments so you can read or listen without wading through a full episode. “If you can’t listen to the whole podcast and you don’t want to read a summary, this is the best way to just get an idea of what’s happening,” Beykpour said.
- Runs a dedicated podcast ad search engine. This one’s clever. Radar can find every episode where a given company advertised and chart how that spend trends over time. It also tracks listener ratings, chart rankings, audience size estimates, sponsorship data, political bias analysis, and brand suitability.
The idea grew out of Particle’s own news-reading app, which used the same tech to drop relevant podcast clips next to related stories. The team saw the feature was valuable but stuck inside the reader. As momentum around AI agents picked up, they spun the podcast intelligence into a standalone API and pivoted the company toward it.
And the API is the point. Radar has a clean web interface, but Beykpour is clear that the real product is the API and MCP, which let AI agents and businesses tap the same intelligence programmatically. That’s why the customer list skews toward hedge funds, AI search platforms, and data resellers. Exa, the search API provider for AI agents, is already a Radar partner. Journalists and researchers can use it too, but the enterprise buyers are the ones writing the big checks.
Pricing, per TechCrunch AI:
- $29 a month per seat for individuals.
- $399 a month for businesses, which bundles 20 seats.
- Custom pricing for API users, based on usage.
Worth naming the limits. Right now Radar is podcasts only. No YouTube, no news clips, no general audio yet, though the company says support for those is on the roadmap. And an index this size lives or dies on transcription accuracy and how well the entity recognition holds up across 135 wildly different verticals, from finance to true crime.
Why it matters: as more of the internet gets consumed and acted on by AI agents rather than people, the formats those agents can’t read become valuable precisely because they’re locked. Podcasts are a huge, growing, mostly untranscribed pile of primary-source commentary. Radar is trying to be the company that turns all of it into structured, queryable data. If audio really is the next frontier agents need to see, indexing it first is a strong place to stand. More details are available at the original TechCrunch AI report.