Image credits: Dexaai
If you listen to a lot of podcasts, you might remember some interesting tidbits and wonder about them. “Wait, who talked about eating fries with sriracha again?” or a more serious question. To find the answer, you must first find the podcast and then search its transcript. Dexa is leveraging her AI to make finding podcasts easier.
This tool allows you to ask about a single podcast, such as Andrew Humberman's Huberman Lab podcast in the screenshot below, or query all podcasts in Dexa's database. We currently have over 120 podcasts and more are being added. Search results provide a summary of the AI-generated answers and pointers to podcasts where participants discussed the topic.
For example, you can ask questions like, “What is the best way to get more sleep?” Then, find answers with time-stamped links to those conversations in Dexa's podcast library. You can also @mention specific podcasts to narrow down your search results.
When I asked Dexa, “What does Mark Zuckerberg think about threads?” I got back a conversation on Lex Friedman's show about The Social Network. But it also showed results with the podcast host talking about Mark Zuckerberg's comments on a thread on another platform.
These searches and results can also be shared with others through links. Because of this, Dexa can also index some of his content for search engines such as Google.
company
Dexa was founded in 2023 by Riley Tomasek. He previously co-founded an app called Flight that allows users to capture notes, images, and ideas on different boards and collaborate across them. The company was acquired by Figma in 2021.
Tomasek said he has become really healthy. As part of the knowledge building phase, he listened to his Huberman Labs podcast. He wanted to take the supplement Huberman suggested, but he forgot the dosage. So he started looking back at when neuroscientists had talked about it, a process that took a long time.
“After this experience, I decided to build something that would make this easier because I never wanted to go through this process again. So I transcribed every episode and used tools like OpenAI embeds and Whisper. to make the search work and posted it to X,” Tomasek said.
After Tomasek posted about his tool, Huberman retweeted it, creating inbound interest in developing the tool from podcasters and listeners. He said a lot of podcasters talk about people not being able to find certain content.
The startup uses a variety of indexing techniques combined with a knowledge graph of people, episodes, and shows to power searches. Tomasek said the tool launched about nine months ago and has had 50,000 visitors to the site each month and answered more than 1 million queries.
Dexa has raised $6 million in seed funding led by Abstract Ventures and The General Partnership with participation from Maple VC and Guillermo Rauch. The company currently has three people on its team and plans to hire more engineering staff in the future.
The company is launching a redesigned site and announcing a partnership with Huberman Labs that will see Dexa search embedded into the podcast site. To do so, the startup trained a special model. Think of this as his GPT for ChatGPT.
The new design allows users to ask follow-up questions and watch linked conversations without leaving the window.
Future roadmap
Dexa's growth has been mostly organic, with podcasters introducing people to the tool. He said the approach still has momentum because the company is working with popular podcasters. Therefore, we are not considering investing in outreach at this time.
“We briefly experimented with advertising on podcasts, but our experiment ended up with a relationship where we indexed the content, added it to Dexa, and people started talking about it naturally. Okay,” he said.
The company is currently not making a profit, but is considering both subscriptions and advertising as sources of revenue as it expands its business.
Dexa also plans to build a mobile app that can increase podcast consumption through short clips. Additionally, the company plans to introduce account creation that will allow users to save search results.
opportunities and challenges
There are many tools to help podcasters transcribe, including Riverside, Podcastle, and MusixMatch. On the search side, the newly pivoted Very Disco is trying to become a sort of podcast IMDB, with information about books, people, TV shows, and more that are mentioned in the podcast.
Marco Paglia, chief product officer at MusixMatch, sees promise in what he calls Dexa's “assistant-driven UX.” Additionally, he feels there is potential to get specific people's opinions through podcasts. But the reason he doesn't embed such searches into his MusixMatch is because he feels text is not the best way to interact with his LLM.
“We feel that typing in chat is not the best UX for Assistant. In fact, we strongly believe that voice is the future for this use case,” he said.
Ben Cmejla, partner at The General Partnership, believes Dexa's approach to harnessing creator knowledge for search is unique.
“A platform's success or failure depends on the trust it builds with its users and content partners. Dexa's approach puts trust first. It’s about trust and trust from our partners that their content will be seen and not just used for training data and anonymized summaries,” he said.