Particle pivots from AI newsreader to podcast intelligence API with Radar

Particle, the AI newsreader startup created by former Twitter engineers, is moving beyond app-based news delivery and into podcast intelligence, unveiling Radar on Wednesday as a searchable podcast engine built for AI agents, researchers and data buyers. The service matters because it turns spoken conversations into structured, machine-readable data that software can search, summarize, monitor and act on.

The pivot reflects a broader shift in the AI market: value is increasingly moving from consumer-facing interfaces to infrastructure that can feed agents, analytics tools and enterprise workflows. Particle says Radar already indexes more than 130,000 podcasts and is drawing especially strong interest from hedge funds, AI search companies and data resellers.

At its core, Radar is designed to do something most internet tools still struggle with: make audio as usable as text. That includes transcribing episodes, identifying speakers, extracting quotes, tagging entities such as people and brands, tracking mentions over time and delivering alerts through email, Slack or webhooks. For companies building AI systems, that means podcast conversations can be queried and operationalized rather than simply listened to.

Why Particle is betting on podcast intelligence

Particle’s shift into podcast search is less of a side project than a strategic realignment. The company originally built a news-reading app that surfaced podcast clips alongside related stories, but leadership eventually saw that the underlying podcast data had far more commercial value than the consumer feed in which it was hidden.

Radar emerges from that insight. Instead of being just a feature that enriches a news app, podcast intelligence is now the product itself, packaged as an API and an MCP service that other businesses can integrate directly into their own systems.

Particle co-founder and CEO Sara Beykpour said the company’s goal is to build a single intelligence layer for audio, noting that most AI agents are built around web text and remain effectively blind to spoken media unless it has already been transcribed.

That gap is the opening Particle hopes to exploit. If agents can search websites, documents and databases, Radar is meant to give them equivalent access to the vast amount of information locked inside podcasts and other audio programs.

How does Radar work?

Radar works by combining large-scale transcription with semantic understanding. It does not simply convert speech to text; it also identifies what the conversation is about, who is being discussed and why a particular segment may be relevant.

The result is a system that can turn an episode into structured data. Users can search for a person, company, product or topic, then receive the exact timestamped segment where that mention appears. The platform also generates clip-based highlights, allowing users to quickly listen to the most relevant excerpts without scanning a full episode.

What makes the search different?

Radar’s search is built around context, not just keywords. That means it can locate material based on relationships between entities and topics, as well as on the content of the spoken remarks themselves.

The product also supports alerts and digests, giving users real-time or periodic notifications when a chosen entity comes up in a podcast. Those alerts can be sent to email, Slack or webhook endpoints, which makes the system useful inside newsroom workflows, investment research pipelines and automated monitoring tools.

• Searches across more than 130,000 podcasts
• Indexes 20,000 new episodes per day
• Includes all Apple Top 200 podcasts across 135 verticals
• Provides speaker labels and rich metadata
• Offers alerts via email, Slack and webhooks

Who wants this data most?

Radar’s earliest commercial traction appears to be coming from customers that need niche, timely information that traditional web search cannot easily capture. According to Beykpour, hedge funds have become the startup’s highest-volume direct API customers, suggesting strong demand for alternative data sources that can support investment research and automated monitoring.

That customer base makes sense. Podcasts often feature executives, founders, political figures and subject-matter experts speaking in a less scripted environment than press releases or official websites. For investors, that can mean early hints about strategy, sentiment, spending, competition or regulatory risk.

Beyond finance, Particle says journalists, researchers, AI search platforms and data resellers are also using or evaluating the product. One partner named by the company is Exa, which provides search infrastructure for AI agents and can incorporate external intelligence into retrieval workflows.

Beykpour told TechCrunch that the strongest interest has come from hedge funds using the API directly, while other paying customers include AI search companies and data resellers.

What data can Radar expose?

Radar goes beyond basic transcription to surface the surrounding ecosystem of each podcast episode. The system can track not just who spoke, but what topics were discussed, which brands or products were referenced, and how those references evolve over time.

It also covers adjacent signals that matter to commercial users, including episode ads, listener ratings and reviews, audience estimates, sponsorship data and political bias analysis. Particle says this broader layer of intelligence creates more ways to monetize the platform, especially for companies that need media monitoring rather than simple playback or transcription.

Podcast ads as a search category

One of the more unusual features is a dedicated search engine for podcast advertising. Users can look up episodes in which a particular company ran ads, then track how that activity changes over time.

That capability is relevant for media buyers, brand teams and competitive intelligence specialists, since podcast sponsorships can reveal how advertisers are allocating budgets across shows and audiences. It also opens another revenue stream for Particle beyond standard search and transcription.

How big is the podcast index?

Radar’s scale is central to its pitch. Particle says the system covers more than 130,000 podcasts, a number it describes as the largest transcribed podcast collection currently available.

The index includes every podcast in Apple’s Top 200 across 135 categories, and the company says it is adding around 20,000 episodes each day. That constant refresh matters because podcast relevance often depends on recency, especially in news, politics, technology and financial markets.

Why AI agents are central to the product

Radar is being positioned less as a standalone consumer tool and more as a data backend for AI systems. That matters because the rise of agentic software has created demand for tools that can fetch information from less accessible sources and present it in a usable form.

Most agents are already good at crawling websites and reading text-based documents. Audio has remained a major blind spot. Podcasts contain interviews, analysis, commentary and breaking news that often never appears in text form, or appears only much later in summarized form. Radar is designed to close that gap.

The company’s API and MCP support make that integration easier. In practice, that means another app, search layer or automated agent can query Radar programmatically, pull relevant segments and use the results in downstream workflows.

What is the MCP layer for?

The MCP layer helps AI systems connect to Radar as a tool rather than as a website. That gives other products a standardized way to request results, retrieve clips and incorporate audio intelligence into larger agent workflows.

For the market, this is important because software vendors increasingly want data sources that can plug directly into model-driven applications without manual intervention. Radar is trying to become one of those sources for audio.

How much does Radar cost?

Radar uses a mixed pricing model that combines self-serve subscriptions with custom enterprise and API pricing. The entry-level plan costs $29 per month per seat, while a business plan is listed at $399 per month and includes 20 seats. Customers using the API are quoted according to their needs.

This structure suggests Particle is targeting both small teams that want a practical listening and search tool, and larger organizations that care more about integration, automation and scale.

What does the pivot say about the AI market?

Particle’s move is another sign that the AI economy is consolidating around infrastructure, not just end-user products. Many startups that began with a consumer front end are now finding stronger economics in APIs, data products and services that can be embedded inside other applications.

That trend is particularly visible in media intelligence. News, podcasts, video and social content are no longer valuable only as content streams; they are also becoming training, retrieval and monitoring inputs for AI systems. The company’s strategy suggests that whoever can reliably index and structure those inputs may own a durable business layer in the age of agents.

It also reflects the growing appetite for alternative data. Financial firms, analytics vendors and search providers want sources that are timelier and less commoditized than public web pages. Spoken media can offer that edge because it often captures current thinking before it is formalized elsewhere.

How could Radar expand next?

Particle says podcasts are only the beginning. The company plans to extend Radar’s capabilities to other audio formats, including YouTube videos and news clips.

If successful, that expansion would widen the system’s reach from a podcast intelligence tool into a broader audio search and monitoring platform. YouTube in particular would dramatically increase the amount of searchable spoken content, though it would also raise the technical and rights-related complexity of building a comprehensive index.

Future growth may depend on whether Particle can keep improving the accuracy, speed and usefulness of its transcription and entity detection while staying ahead of bigger platform players. For now, the company is betting that audio intelligence is still an under-served market with clear demand.

Timeline of Particle’s move into Radar

The company’s shift can be understood as a progression from consumer feature to infrastructure product.

What happens next for Particle?

Particle now faces the challenge many AI startups encounter after a promising pivot: proving that technical sophistication translates into repeatable revenue. The company has identified a real pain point, built a sizable index and found an early customer segment with clear willingness to pay.

The next test is whether Radar can become essential infrastructure for enough businesses to sustain long-term growth. If it does, Particle may end up being remembered less as an AI newsreader startup and more as one of the companies that helped make audio legible to machines.

For now, Radar gives Particle a clearer value proposition: not just helping people consume media, but helping AI systems understand what people are saying when the news is being made in their own voices.