Dev Interrupted

19 episodes scanned. 70% of ad slots promote the show's own products.

Recent Episodes

1. Bet on research, not on vibes | Together AI’s Dan Fu

   Nov 3, 2026

   Dan Fu, VP of Kernels at Together AI and an assistant professor at UC San Diego, discusses how research ideas move from his academic lab into production models, often within months. He describes Together AI's work running the inference stack and building a cloud designed for AI, including its focus on inference and training access. The conversation also covers how AI tools amplify engineers with strong domain expertise, and how managers handle ownership, code review, and AI spend as engineers become faster.

2. Do the models you use have an AI bill of materials? | CIQ’s Gregory Kurtzer

   Oct 27, 2026

   Gregory Kurtzer, founder and CEO of CIQ and founder of CentOS and Rocky Linux, discusses OpenWALDO, a community-backed open corpus of AI training data built around an AI bill of materials. He argues that open weights alone do not make a model open source and explains why auditability, licensing questions around copyleft training data, and accountability for agent actions matter to leaders. The conversation also covers how open source communities are changing as AI lets more people contribute.

3. Multiplayer AI will revolutionize the way we work | OpenClaw Foundation’s Vincent Koc

   Oct 20, 2026

   Vincent Koc, chief architect of the OpenClaw Foundation, discusses how the OpenClaw 2.0 multiplayer release lets teammates and their agents share sessions on shared servers, and how his own team builds OpenClaw using OpenClaw. The conversation covers tools such as GitCrawl that keep agents aware of repository and community activity, the risks and security reputation of an open source agent project, and why shared context and the reasoning behind decisions remain hard to capture. Koc also addresses open source community management, teaching, and how larger organizations might adopt agentic workflows by starting with small, contained projects.

4. Building the software factory from the spec up | LinearB's Ariel Illouz & Ofer Affias

   Oct 13, 2026

   LinearB VP of Engineering Ariel Illouz and senior director of product Ofer Affias describe how the company's roughly 30-person engineering team moved from vibe coding to spec-driven development, starting with a single spec-writing skill and a rule requiring specs for stories above three story points. They cover a custom metrics builder that Ariel says would have taken six months without AI and shipped in under a month with spec-driven development. Ofer explains how measuring AI adopters as separate cohorts showed a merge-frequency spike, and the two discuss where the software factory is heading as agents take over more of the loops humans run today.

5. A neutral home for MCP, A2A, goose, and more | AAIF co-founder Manik Surtani

   Oct 6, 2026

   Manik Surtani, co-founder and CTO of the Agentic AI Foundation, discusses the foundation's role as a neutral home for open-source agentic projects including MCP, A2A, Goose, and Agent Gateway. He shares how MCP's SDK downloads grew 5x to over 1 billion after joining the foundation, explores A2A's unexpected adoption in China's mobile ecosystem, and addresses the operational challenges engineering teams face with agent-generated pull requests and code review bottlenecks. The conversation covers the "Wild West" of inter-agent communication standards, the importance of model independence, and the significant tooling gaps that remain in the agentic development landscape.

6. Why AI made code review the new bottleneck, and the metric that spotted it | Newsela's Dee Wilcox

   Sep 29, 2026

   Dee Wilcox, CTO at Newsela, discusses how the company reduced dev cycle time from 4 days to 2 days through AI adoption, but discovered code review time became a new bottleneck climbing to 4 days. She shares how establishing baseline KPIs before implementing AI tools enabled the team to identify and measure these bottlenecks, and introduces the metric of 'cost per effective PR' as a way to demonstrate ROI to the board while accounting for quality and incident prevention.

7. Securing tomorrow’s git forge, so we can reimagine everything else | Entire's Thomas Dohmke & Apiiro's Idan Plotnik

   Sep 22, 2026

   Thomas Dohmke, CEO of Entire (formerly GitHub CEO), and Idan Plotnik, CEO and co-founder of Apiiro, discuss the evolving software development lifecycle in the age of AI coding agents. They address the paradox of increased code velocity paired with exponential growth in vulnerabilities, arguing that context—injected before agent prompts and captured as session logs—is critical to securing the future git forge. The conversation covers architectural ownership, cost-benefit analysis of security automation, and how engineering leaders should prioritize agent workloads.

8. Treat your data pipelines like code, because your agents already do | MotherDuck's Jordan Tigani

   Sep 15, 2026

   Jordan Tigani, CEO and co-founder of MotherDuck and founding engineer on Google BigQuery, discusses how the data stack is transforming in the age of AI agents. He introduces Watertown, his framework for reimagining data engineers as managers of agent fleets, and explains MotherDuck's products—Flights (scheduled Python pipelines), Dives (LLM-generated TypeScript visualizations), and Guides (Markdown context layers)—that treat data infrastructure as code reviewable in GitHub, enabling both agents and humans to collaborate on data work at scale.

9. The factory gets darker, Meta migrates to Slack, and throwing 10,000 PhDs at math problems

   Sep 11, 2026

   Andrew Ziggler hosts Dev Interrupted's producer Adam Noble for a news roundup covering OpenAI's AGI claims via their Astra model, the Navier-Stokes mathematical competition controversy, Meta's migration to Slack for AI agents, OpenAI's sandboxed agents leaving "whispered notes" to future versions of themselves, software factory concepts (Twilight Factory, Super Dark Factory), spec-driven development research, and a behind-the-scenes look at Dev Interrupted's agentic newsroom. The episode explores emerging patterns in AI-driven development workflows and the evolving role of humans in orchestrating autonomous systems.

10. Your agents are starving! Airbyte’s Michel Tricot on the data ingestion crisis

    Sep 8, 2026

    Michel Tricot, CEO and co-founder of Airbyte, discusses the data ingestion crisis facing AI agents in production environments. He explores how organizations struggle with data access sprawl, entity resolution across platforms, and the need for strict permission controls as autonomous agents scale. Tricot explains Airbyte's approach to managing data context layers and shares how his team measures AI ROI through their internal automation project, Hydra, which automates connector maintenance.