Language Model Builder with Felix Rieseberg

A young man who's a sophomore in high school emailed me to come on his podcast that he does with his friend, and I had so much fun. I decided to make it a crossover episode. Check out "Compile after Class" on this week's episode! https://www.youtube.com/@shreemanickaraja

Maynard Holliday spent 30+ years at the intersection of robotics, nuclear security, and defense technology, most recently serving as the Assistant Secretary of Defense for Critical Technologies. He's worked on everything from helping remediate Three Mile Island to securing tons of Soviet nuclear material in post-Cold War Ukraine to overseeing the Pentagon's critical technology portfolio including AI, hypersonics, and microelectronics. He shares what "trusted AI" really means, why he co-founded Black in Robotics, and how T'Challa and Tony Stark shaped his career more than you'd expect. https://www.maynardholliday.com/

Felix Rieseberg came to computing through poetry, and that background gives him a unique lens for explaining language models. A longtime engineer, Felix now gets to watch AI evolve from the inside. In this episode, Scott and Felix explore why language models are surprisingly easy to understand despite feeling magical, what it's like to train a model from noise to coherence, and the bigger philosophical questions about intelligence that AI keeps surfacing. You can learn how AI really works by making your own in a weekend with his free https://languagemodelbuilder.com/

Varun Badhwar, CEO and Co-Founder of Endor Labs, joins Scott to explore the expanding software supply chain risk in the age of AI-generated code. They discuss how developers have gone from writing code to assembling it from layers of dependencies like open source, third-party libraries, and now the outputs of AI...and what all this means for security. From hallucinated npm packages that could be squatted on, to developers becoming part of the attack surface themselves, this is a wake-up call for anyone building software today. https://www.endorlabs.com/

Alexandra Lato, VP of Product at Stack Overflow, joins Scott to talk about how Stack Overflow is evolving for the age of AI agents. They explore the new "Stack Overflow for Agents" feature, where a single prompt bootstraps an agent with all the knowledge and APIs of Stack Overflow, and what it means for how developers, agents, and communities share and verify knowledge together. https://agents.stackoverflow.com/

Dr. Chinasa T. Okolo is the Founder and Scientific Director of Technecultura and a consultant for the United Nations and the World Bank. She talks with Scott about the yawning gap between AI's pace of development and policymakers' ability to understand it. She also discusses what it means to do AI governance research with a focus on Global Majority communities that are too often left out of the conversation. https://www.chinasatokolo.com

Scott talks with Nathan Sobo, CEO and co-founder of Zed, about what comes after the traditional code editor. They start with Zed’s vision for a fast, collaborative, AI-native development environment, then go deeper on DeltaDB: a new approach to versioning software at the operation level, not just at the commit level. Nathan explains why so much important software work happens “between commits,” how agent conversations and code changes can become durable shared artifacts, and what it might mean for Git, collaboration, and the future of programming tools. Nathan previously helped build Atom at GitHub, and Zed describes DeltaDB as operation-level version control for human and AI collaboration. https://zed.dev/deltadb

On this episode of Hanselminutes, Scott talks with Ed Rogers of Bristol Braille Technology about the Canute project and the long road toward affordable multiline Braille. Most refreshable Braille displays show a single line at a time; Canute changes the experience by giving readers nine lines and 360 cells of spatial context. Ed shares how multiline Braille opens up new possibilities for reading, coding, math, music, diagrams, education, and independence and why Braille remains a vital technology for literacy, employment, and full participation in the digital world. https://bristolbraille.org

Charity Majors is the co-founder and CTO of Honeycomb.io, where she pioneered the concept of modern observability for distributed systems. Before Honeycomb, she spent years at Parse (acquired by Facebook), Facebook, and Linden Lab building Second Life. She is the co-author of "Observability Engineering" and "Database Reliability Engineering" (O'Reilly). She writes at charity.wtf. Whatever she and Scott are planning, they are gonna observability the heck out of it.

Scott talks with Aran Khanna, co-founder and CEO of Archera, about a new category of cloud financial tooling: "Insured Commitments." Instead of locking into 1- or 3-year reserved instance contracts and hoping your usage matches, Archera offers commitments as short as 30 days. They get into the economics of cloud purchasing, how AI workloads are changing capacity planning, and what FinOps looks like in 2026. http://archera.ai