As AI makes software development significantly easier, traditional business moats are eroding. This episode argues that real-world, human-generated data is becoming the only reliable moat for software businesses in the near future. Founders should focus on collecting, validating, and making this unique data accessible, as exemplified by PodScan's success with 50 million transcribed podcast episodes.
Summarized by Podsumo
AI's Impact on Software Moats: The increasing ease of building software with AI tools is diminishing traditional moats like development difficulty and maintenance, shifting the focus to other competitive advantages.
The Bifurcation of Data: Data is splitting into highly valuable human-generated data (signal) and increasingly commoditized AI-generated data (slop), with human data being inherently more valuable due to its origin and exclusivity.
Data as the Sole Moat: Real-world, human-generated, validated, and cleaned data is identified as the primary and most reliable moat for software founders, as AI cannot replicate its unique creation process.
The Importance of Data Accessibility: Beyond just collecting unique data, making it accessible—especially through an API-first approach and ensuring UI/API parity—is crucial for maximizing its value and enabling automation for users and AI agents.
Metadata as a Unique Data Source: Even incidental metadata collected from user interactions (e.g., posting times, engagement patterns) can form a unique and valuable data moat that competitors cannot easily replicate.
"I believe that good data, and that is real world data, mostly human generated, it's validated and it's cleaned, is the only reliable mode that we have as software founders in the near and midterm future."
— Arvid Kahl
"Human generated data is valuable just by the sheer fact that it's not AI generated at this point."
— Arvid Kahl
"Having data is half the mode. A veiling data is the other half."
— Arvid Kahl