↘ Talks Archive
Conversations on technology & the economics of software.
// Archive_Metadata
Count: 38 episodes
Status: Active
Focus: tech_moats, ai_infra, venture_economy
Latest Release
We are in a Race To Understand AI | Eric Ho, Goodfire
Goodfire co-founder and CEO Eric Ho joins Lightspeed partner Nnamdi Iregbulem to explain why AI interpretability is becoming critical infrastructure, and how opening the black box of neural networks can enable safer deployment, intentional model design, and deeper scientific discovery.
“Inside the Black Box: The Urgency of AI Interpretability”
Listen now →
Episode Summary
Recorded live at Lightspeed’s offices in San Francisco, this special episode of Generative Now dives into the urgency and promise of AI interpretability. Lightspeed partner Nnamdi Iregbulem spoke with Anthropic researcher Jack Lindsey and Goodfire co-founder and Chief Scientist Tom McGrath, who previously co-founded Google DeepMind’s interpretability team. They discuss opening the black box of modern AI models in order to understand their reliability and spot real-world safety concerns, in order to build AI systems of the future that we can trust.
Delta Podcast Ep. 21: Nnamdi Iregbulem, Lightspeed Partner
Delta Institute (deltainstitutes.org) is the most talent-dense community of engineers and researchers in the world. They have hosted events across the Bay Area, including both Stanford and Berkeley. Past event co-hosts include partners at Sequoia and Andreessen Horowitz, and one of the first ten employees at Cursor. Delta Institute also hosts the Delta Podcast, a series of video interviews with top researchers/engineers in order to help share their experiences and reach a broader international audience.
Benedikt Stroebl: With Imperfect Verifiers, Scale Fails
Verifiers are a hot topic in AI these days, where they play a role in both post-training and inference time scaling. But what if more compute at inference time doesn't always improve AI performance? I discuss with Benedikt Stroebl his research on the challenges presented by the inference time scaling paradigm, along with a particular application of agentic techniques to the world of cybersecurity.
Satish Chandra: Google's AI Coding Agents That Fix Bugs Automatically
Can AI agents soon eliminate the tedious task of debugging code? Google's Satish Chandra reveals groundbreaking research on how AI is changing the way teams fix and maintain production code.
Jonas Hübotter: 1000 Expert Models, 100x Faster, For Free
What if you could get all the benefits of test-time adaptation of AI models for a particular task, but 100 times faster and essentially for free? Jonas Hübotter shows how pre-training 1000 expert models and merging them at test-time achieves exactly that, with nearly zero overhead.
Daniel Kang: AI Hackers Are Coming Sooner Than You Think
AI systems that can autonomously hack applications are no longer science fiction. In this revealing conversation, Daniel Kang breaks down his team's research showing that AI systems can now find and exploit security vulnerabilities using agentic approaches.
Andy Zhou: Language Agent Tree Search & Autonomous Red Teaming
In this conversation I chat with Andy Zhou, where we discuss his ICML-accepted work on Language Agent Tree Search, which unifies reasoning, acting, and planning in LLMs via an innovative search algorithm, and AutoRed Teamer, an autonomous system for end-to-end red teaming LLMs.
Charlie Snell: Optimally Scaling Test-Time Compute & Predicting Emergence
In this conversation, I sit down with AI researcher Charlie Snell from UC Berkeley to discuss his groundbreaking work on optimally scaling test-time compute and predicting emergent capabilities in large language models. Charlie unpacks multiple dimensions of his work that could reshape how we approach AI development and inference strategies.
Developer Productivity, Real-Time Data Infrastructure, and The Fat-Tailed Nature of Enterprise Software – Datacast
The 109th episode of Datacast is my conversation with Nnamdi Iregbulem, a Partner at Lightspeed Venture Partners. His mission is to increase total software output by supporting entrepreneurs building technical tools for technical people. He focuses on investments in technical enterprise software such as developer tools, application infrastructure, and machine learning.
View Full Archive
Browse all 38 recorded appearances including panels, guest lectures, and podcasts.
Join the Syndicate
Long-form essays on the economics of tech, with the data to back it up. All signal, no noise.
Read by 5,000+ founders, engineers, and investors.