Episodes

  1. EP 020 Ben Morss Why AI Can Generate a Song but Struggles with Music Theory with Ben Morss · Oct 13, 2026 · 01:09:03 Dan Gerlanc and Mistral's Ben Morss discuss the gap between AI output that works and work people actually find valuable. A sing-along app that broke live and models that struggle with music theory show why understanding a system and judging its output still matter.
  2. EP 019 Cory O'Daniel Agents, Engineers, and Citizen Developers with Cory O'Daniel · Oct 6, 2026 · 01:17:01 Dan and Cory O'Daniel discuss how infrastructure agents gain safe autonomy through modeled context, separate identities, and enforceable permissions. Cory shares Massdriver's development and testing workflows and argues that engineers should build guardrails for citizen developers.
  3. EP 018 Stephanie Jarmak Building a Software Factory with Gas City with Stephanie Jarmak · Sep 29, 2026 · 01:19:28 Dan and Stephanie Jarmak discuss how she became a Gas City maintainer by turning agent review feedback into an automated contribution workflow. Her path from planetary science to AI engineering reveals both the freedom to build and the reliability problems that still demand human judgment.
  4. EP 017 Eric Ma Agentic Data Science in a Reproducible Marimo Notebook with Eric Ma · Sep 22, 2026 · 58:17 Eric Ma shows how Marimo Pair turns a notebook into a shared workspace for data science agents, using a protein-engineering analysis to explore reproducibility, visualization, and the limits of automation. He and Dan also discuss design guardrails, opinionated repository structure, and a one-foreground-task practice for managing agentic work without burning out.
  5. EP 016 Andrew Zigler Beads, Better Specs, and Less Rework with Andrew Zigler · Sep 15, 2026 · 01:07:48 Dan and Andrew Zigler explore how upfront alignment and Beads task graphs can reduce rework in agentic development. They discuss shared team memory, Andrew’s personal agent setup, and the changing role of testing and code review.
  6. EP 015 Julien Verlaguet Programming Languages for AI Agents with Julien Verlaguet · Sep 8, 2026 · 01:18:41 Julien Verlaguet explains how SkipLabs is building Skipper, a closed-loop coding agent around an AI-tailored, reactive development environment. He and Dan Gerlanc weigh the promise of faster software work against code quality, junior-engineer training, and the limits of formal verification.
  7. EP 014 Thomas Wiecki AI Agents for Bayesian Data Science with Thomas Wiecki · Sep 1, 2026 · 01:14:00 Thomas Wiecki and Dan Gerlanc explore how Bayesian models become more useful when agents can use them, from scenario simulation to team-chat data science. They also examine verification, lightweight workflows, institutional knowledge, and the responsibility that comes with making advanced methods broadly accessible.
  8. EP 013 Hugo Bowne-AndersonDoug Turnbull Search is Eating AI with Hugo Bowne-Anderson, Doug Turnbull · Aug 25, 2026 · 41:17 Hugo Bowne-Anderson and Doug Turnbull join Dan to explain why agentic search begins with retrieval, measurement, and clear product goals. They then examine what abundant AI-generated code demands from builders: deeper judgment, stronger verification, more restraint, and interfaces that preserve human attention.
  9. EP 012 Paul Everitt The Handoff Packet for Local Agentic AI with Paul Everitt · Aug 18, 2026 · 01:05:37 Paul Everitt and Dan Gerlanc explore how static analysis, structured tool calls, spec-driven development, and tightly scoped handoff packets can make local agentic coding more reliable. They discuss hybrid model architectures, the roles of large, medium, and small models, and why developers still need to read, test, and own the code.
  10. EP 011 Niels Bantilan How AI Agents Change the Work of an ML Engineer with Niels Bantilan · Aug 11, 2026 · 01:01:13 Dan and Niels Bantilan discuss how agents are changing ML engineering, from threefold coding velocity and Kubernetes debugging to narrowly scoped production workflows. They also explore Flyte as an AI runtime and Pandera's future in multimodal data validation, while weighing the risk of losing deep understanding when agents explain and fix problems.
  11. EP 010 Jacob Young Reducing Entropy in Agentic Software with Jacob Young · Aug 4, 2026 · 01:06:44 Jacob Young reflects on his experience doing technical due diligence in the age of agentic engineering. We discuss the commonalities in teams that succeed with agentic engineering and tips for battling entropy as the cost to generate code drops.
  12. EP 009 John Berryman Unharness Your Agents with John Berryman · Jul 28, 2026 · 01:05:40 Dan and John Berryman explore what happens when agents move beyond terminal coding into websites, applications, physical spaces, and everyday workflows. They discuss Rook, English-language skills, the limits of current memory systems, and the safety standards needed before agents can act broadly on a person’s behalf.
  13. EP 008 Sheamus McGovern The AI Skill Flip with Sheamus McGovern · Jul 21, 2026 · 01:03:58 Sheamus McGovern, founder of ODSC and author of The AI Skill Flip, argues that AI is changing which skills matter rather than eliminating jobs, and that passion, creativity, and judgment separate the people getting real value from AI from everyone else. He and Dan dig into evaluation, prompt and context engineering, the AI technical debt piling up behind vibe coding, and why junior engineers still need the classical skill set.
  14. EP 007 Matthew Rocklin The New Sport of Programming with Matthew Rocklin · Jul 14, 2026 · 55:44 Matthew Rocklin explains how agents help him run a highly automated version of his company and why broad context across business systems can reveal problems that siloed tools miss. He and Dan also discuss the new mental habits, verification systems, open source trade-offs, and personal ambition required when agents can build most of the code.
  15. EP 006 Jonathan Bown Skills, Context, and Trust: The New Agentic Coding Stack with Jonathan Bown · Jul 7, 2026 · 01:17:15 Jonathan Bown, Principal ML Engineer at Western Governors University, and his team apply evaluation-driven development with MLflow 3.0 to iterate a pre-enrollment agent from nearly being cancelled due to negative feedback into one hitting roughly 82 percent correctness. They also dig into the new landscape of MLflow and Databricks skills, agentic coding tools like Kiro and Antigravity, and where Jonathan still draws hard lines around trust.
  16. EP 005 Greg Ceccarelli When Software Gets Cheap, Focus Gets Expensive with Greg Ceccarelli · Jul 2, 2026 · 01:07:43 Dan and Greg discuss the evolution of agentic engineering and the biggest challenges teams face as they become agent-native. Learn about Greg's new book, "25 Patterns in Agentic Engineering," Greg's preference for local agents, and why writing copy and judging product quality remain difficult to delegate.
  17. EP 004 Eleanor Berger From Supervising AI to Building Systems for It with Eleanor Berger · Jun 18, 2026 · 59:51 From closely supervising LLMs to building systems that let async, cloud-based agents work unattended, Eleanor explains how her work has changed over the last six months. We discuss the primitives of an agent system, why deterministic verification is a bottleneck, and how software development is becoming a sub-branch of systems engineering.
  18. EP 003 Angie JonesDemetrios Brinkman Are We All Managers Now? with Angie Jones, Demetrios Brinkman · Jun 4, 2026 · 01:04:53 Dan, Angie, and Demetrios explore how the latest frontier models have turned software engineering into something closer to managing a team of agents, and why governance, tokenomics, and design now matter more than typing code.
  19. EP 002 Ian Stokes-Rees Claude-maxxing: Burning $10K in tokens for only $50 with a custom software factory with Ian Stokes-Rees · May 21, 2026 · 01:07:42 Learn how Ian runs six to ten Claude Code agents as a disciplined Scrum team with up to 100 PRs and billions of tokens in a single day. Hitting Claude Max 20x limits, token-management tricks, what to trust (and not trust) from agentic engineering.
  20. EP 001 Greg Sadetsky Vibe Coding in the Physical World: Robotics, Circuits, and Dangerous Permissions with Greg Sadetsky · May 9, 2026 · 01:12:27 Greg Sadetsky on driving laptop-refurbishing robots with AI vision, the near-100x speedups he sees in well-tested code — and what AI still can't do, including vibe-coding physics.