Thomson Reuters
- A version of this story originally appeared in the BI Tech Memo newsletter.
- Sign up for the weekly BI Tech Memo newsletter here.
During a recent interview with venture capitalist Tomasz Tunguz, he mentioned something that really caught my attention.
We were discussing AI coding technology and how these tools are changing the buy-versus-build equation in software. It’s much easier to write software now, so more people are doing it themselves, rather than buying expensive, janky SaaS offerings.
Tunguz, a former Googler, has been investing in software and SaaS startups for almost 2 decades, first at Redpoint Ventures and now at Theory Ventures. So I asked him to put his money where his mouth is: What internal applications has he built himself, using AI coding tools?
The VC said he’s created an AI-powered app that transcribes and summarizes tech podcasts and integrates the information into the workflow of Theory Ventures. He calls it “The Podcast Orchestrator.”
The system finds all of the names of the startups and founders featured in these tech podcasts, figures out who they are and what they’re saying, and then adds the details into Theory’s CRM, a customer relationship management software program that’s commonly used to track and generate sales and related leads.
“End-to-end AI workflows”
It doesn’t stop there, according to Tunguz. This homegrown application also does some of the work of junior VC associates.
“There’s actually an AI associate that is an AI agent that will look at all the related companies, find all the blog posts, look at GitHub, look at Hacker News comments, size of the market, prepare a memo, and then make a recommendation based on our investment criteria,” Tunguz explained. “That’s an example of an end-to-end AI workflows.”
He said he built the application in about two or three hours, using Claude Code for the software writing part and OpenAI‘s open-source audio-to-text AI tool, Whisper, for the transcriptions. “Then I send it to a bunch of different AI systems to summarize the content and find the startups,” he added.
Tunguz runs The Podcast Orchestrator every morning. It takes 5 to 10 minutes to process the new podcast information and then it automatically sends him an email summary and creates tasks in Theory’s Asana project-management software. This helps him quickly go in and review any startups and founders that meet the firm’s investment criteria.
There’s alpha buried in podcasts
The VC created this tool because he ended up listening to 40 or more tech podcasts and it had become difficult to find the time to listen to everything. Instead of just missing out on comments, data, and other valuable nuggets buried in these broadcasts, Tunguz used AI tools to quickly create a solution.
“There’s alpha in these podcasts,” Tunguz said, referring to the extra returns talented investors can generate through an edge or novel insight.
Startup founders, tech executives, and other industry experts often open up more during podcasts — especially the longer they run, Tunguz noted, which makes it more likely there’s valuable information to be mined by his homemade AI tool.
“There are data points in these podcasts that don’t exist anywhere else,” he told me. “And so if we can find those data points and then add them to the right CRM files that’s great. That’s not widely available public information. We can extract that, learn something about it, and then put it into our database and at some point it will be useful.”
A pitch denied
I told him he should release The Podcast Orchestrator as a publicly available app, noting that newsrooms, investment banks, and hedge funds would likely want to use the tool, too. I suggested “Podcast Alpha” as a name for the commercial product.
Tunguz noted that the application at the moment is not ready for public release. It currently runs via a command line on a computer terminal, which means you need some coding chops to fire it up. Turning it into a snazzy app for everyone to use would probably take another couple of hours of AI coding work, he said.
And Tunguz has had formal training as a software engineer, so this would take actual, technical work to pull off.
Still, I pressed him to release it publicly, but he didn’t sound convinced.
“There’s no money in podcasts,” Tunguz said.
Sign up for BI’s Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.
