The archive · Developer & Business Tools · Strategic decision · 2023–2026
Particle's Radar bet: index 130,000 podcasts as the audio layer AI agents can't see
Former Twitter engineers pivot their AI newsreader to Radar: 130,000 podcasts transcribed, sold to hedge funds and AI agents via API.
Particle
What the business is
Audio intelligence platform: Radar transcribes and indexes 130,000 podcasts and serves the data to AI agents and businesses via API and web search.
Starting capital：$4.4M seed (Kindred Ventures, Adverb Ventures), closed April 2023
How it started
In 2023 Sara Beykpour, ex-Twitter senior director of product, and Marcel Molina, ex-Twitter and Tesla engineer, founded Particle as an AI news reader that summarizes stories drawn from sources across the political spectrum. The startup raised a $4.4M seed from Kindred Ventures and Adverb Ventures, closing in April 2023, with angels including Ev Williams and Scott Belsky.
What happened
In February 2026 the app's Podcast Clips feature began surfacing the most interesting moments of podcasts alongside news stories, and it became one of the app's most-loved features. Six months later Particle unbundled the capability into Radar: a podcast search engine and API that transcribes more than 130,000 shows (the largest transcribed index in existence, the company said), adds speaker labels and entity metadata, tracks mentions with alerts, and even indexes podcast ads. Hedge funds became the highest-volume API customers; the search-API company Exa integrated as a partner.
How it ended up
Live: Radar ships consumer search at $29/user/month and a $399/month business plan, with custom pricing for API users; expansion to YouTube videos and news clips is planned.
Background
Particle started in 2023 as an AI news reader built by ex-Twitter engineers Sara Beykpour and Marcel Molina, backed by a $4.4M seed from Kindred Ventures and Adverb Ventures. Its pitch was “understand more, faster”: AI summaries of any story, drawn from sources across the political spectrum. In early 2026 the app launched Podcast Clips, pulling the most interesting moments from podcasts into the news feed — and it became one of the app's most-loved features.
The team realized the intelligence was trapped inside the reader. As AI-agent adoption took off, Particle unbundled it: Radar is a podcast search engine that transcribes and understands more than 130,000 shows — the largest transcribed podcast index in existence — with 20,000 episodes added daily. It carries speaker labels, entity metadata, mention alerts, and a podcast-ads search engine; the real product is the API and MCP server that lets agents and businesses query the audio layer programmatically.
Customers showed up before the launch noise faded: hedge funds are the highest-volume API users, alongside AI search platforms and data resellers, with the search-API company Exa among the partners. Radar is priced at $29 per user per month and $399/month for a 20-seat business plan. The bet is that most agent tooling crawls text, so whoever owns transcribed, structured audio becomes the standard layer for it.
What has to be true
- The pivot starts from a validated feature instead of a hypothesis: Podcast Clips was already used and loved inside the app, so Radar sells something that demonstrably worked.
- Agents and hedge funds both lack an audio layer — Radar's API and MCP integration slot into a genuine gap rather than fighting incumbents for text search.
- Audio is technically hard to index — transcription, speaker separation, entity recognition — so the 130,000-podcast head start is a moat while rivals still license transcripts.
- Paid, differentiated tiers — consumer search, business seats, custom API pricing — let the same technology monetize three very different buyers.
What can be applied
A feature that users love can be worth more as a product than inside the product that birthed it — unbundle the beloved capability and sell it to the customers the app can never reach.
Aftermath
As of early September 2026 Radar is live with paying customers: hedge funds are its highest-volume API users, alongside search platforms and data resellers, and the company's homepage banner already advertises “135,000+ podcasts.” Expansion to YouTube videos and news clips is planned, but the newsreader still exists as the public focus shifts decisively to audio intelligence. Unproven: whether podcast-intelligence data sustains a standalone business — and whether hosts and publishers accept being indexed and resold.
Sources
- Former Twitter engineers are building Particle, an AI-powered news reader, backed by $4.4M
- Particle's AI news app listens to podcasts for interesting clips so you you don't have to
- Radar makes podcasts searchable — and usable by AI agents
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card