Dexa aims to get more out of podcasts with AI-powered search | Tech Crunch

Spread the love


If you listen to a lot of podcasts, chances are you remember the funny tips and wonder… “Wait, who said anything about eating fries with Sriracha again?” Or more serious questions. To search for answers, you first need to find the podcast and then search for their transcripts. Dexa is trying to make podcast search easier by using AI.

The tool allows you to ask questions about a single podcast, such as Andrew Humberman's Huberman Lab podcast in the screenshot below, or query all podcasts in the Dexa database — currently more than 120 have been added. Search results provide you with an AI-generated summary of the answer as well as pointers to podcasts.

For example, you might ask, “What's the best way to get more sleep?” Questions like And find answers from Dexa's podcast library with timestamped links to those conversations. You can also @mention a specific podcast to narrow down your search results.

Image Credits: Dexa

When I asked Dexa “What does Mark Zuckerberg think about threads?” This was the result of his conversation on the Lex Friedman Show about the social network. But it also showed me the results of podcast hosts talking about Mark Zuckerberg's comment about threads on another platform.

You can also share these searches and results with others via links. Because of this, Dexa can also index some content for search engines like Google.

Company

Dexa was started in 2023 by Riley Tomasek, who previously co-founded Flight, an app for users to capture and collaborate across notes, images and ideas on various boards. The company was acquired by Figma in 2021.

Tomasek says he's really gotten healthy. As part of his knowledge building phase, he listened to a Huberman Labs podcast. He wanted to take a supplement prescribed by Huberman but forgot the dosage. So he started looking back when the neuroscientist talked about it, and the process took a long time.

“After this experience, I decided to build something to make it easier because I don't want to go through this process again. So I transcribed all the episodes and used OpenAI's embeddings and tools like Whisper to do the search work and posted it on X,” says Tomasek.

After Tomasek posted about his tools, Huberman retweeted them, and that sparked interest from podcasters and listeners for the development of the tools. He said many podcasters talked about people not being able to find certain content.

The startup uses different indexing techniques combined with a knowledge graph of people, episodes and shows to power its search. Tomasek said the tool was launched about nine months ago and 50,000 people visit the site every month, answering 1 million questions.

Dexa has raised $6 million in seed funding led by Abstract Ventures and The General Partnership in partnership with Maple VC and Guillermo Rauch. The company currently has three people on the team and is now planning to hire more engineering staff.

The company is launching a redesigned site and announced a partnership with Huberman Labs, where Dexa's search podcast will be embedded on the site. For that, the startup trained a special model — think of it as GPT in ChatGPT.

The new design allows users to ask follow-up questions and allows them to see the linked conversation without leaving the window.

A future roadmap

Dexa's growth has been largely organic, with podcasters pointing people toward the tool. The method is still gaining momentum, he said, as the company works with popular podcasters. So not investing in outreach for now.

“We briefly tried advertising on podcasts and found that our efforts culminated in a relationship where we indexed their content and added it to Dexa, and they started talking about it organically,” he said.

The company isn't making any money right now, but considers both subscriptions and ads as sources of revenue as it scales.

Dexa also plans to create a mobile app, where it can provide more consumption for podcasts through short clips. Additionally, the company plans to introduce account creation so that people can save their searches.

Opportunities and challenges

There are many tools that can help podcasters with transcription, including Riverside, Podcastle, and MusicsMatch. In terms of search, the newly pivoted Very Disco is trying to become a sort of IMDB of podcasts, with information such as books, people and TV shows mentioned in the podcasts.

MusixMatch's chief product officer Marco Paglia sees promise in what he calls Dexa's “assistant-driven UX.” Moreover, he thinks there is potential in allowing people to relive a particular person's views through podcasts. But his reason for not incorporating such a search in MusixMatch is that he feels that text is not the best way to interact with LLMs.

“I think typing in chat for assistants is not the best UX. Instead, I strongly believe that voice is the future in this use case,” he said.

Ben Cmejla, a partner at The General Partnership, believes that Dexa's approach to using knowledge from creators for search is unique.

“The success or failure of a platform depends on the trust it builds with users and content partners. Dexa's approach puts trust first — the user trusts that the answers to their questions come from human experts, and they trust that the partner isn't just using their content for training data or anonymous summaries,” he said.



Source link

Leave a Comment