Delta Podcast Ep. 21: Nnamdi Iregbulem, Lightspeed Partner
Nov 28, 2025
Nnamdi Iregbulem, a partner at Lightspeed with an MBA from Stanford, dives into the world of developer tools and AI-native infrastructure. He shares insights on how he self-taught his coding skills and the importance of mechanistic interpretability in AI. Nnamdi discusses Lightspeed's centralized research efforts and the necessity of evaluating AI systems through tools like Patronus. He also offers advice for researchers eyeing startups, emphasizing systems thinking and collaboration. Plus, he reveals his tips on connecting with talent in the tech ecosystem.
AI Snips
Chapters
Transcript
Episode notes
Self-Taught Technical Origin
- Nnamdi taught himself to code as a kid, building websites, modding games, and assembling gaming PCs.
- He studied economics in college and later took CS credits during his MBA, calling most of his CS knowledge self-taught.
Focus On Technical Tooling Below Apps
- Nnamdi focuses on technical tooling: developer tools, application infra, and AI-native layers below the app.
- He views AI adoption as driving the majority of his work, shifting from ~20% to ~90% of his time.
Agents Reshape Infrastructure And Dev Flow
- He is excited about agentic infrastructure and tooling that helps agents access tools and run reliably in production.
- He also highlights broad changes to the software development lifecycle driven by code generation, chats, and agents.