Heavybit Podcasts

Heavybit Podcasts is not one show. It is several, delivered through a single feed. Each series keeps its own hosts, its own framing and its own episode numbering, so the register changes as the feed moves along. The common thread is conversation with people who build software tools, and a running question about what AI is doing to that work. Data Renegades presents itself as a series that interviews pioneers of the tools and concepts the data world now uses every day. Its conversation with Toby Mao makes the case plainly: a data pipeline is software, and the field would be further along if it adopted deployments, checks and testing. Data modeling comes first, and AI only raises the stakes. The coding routine described there is deliberately narrow. The model gets one step at a time, and its code is reviewed as rigorously as a human's, because small errors compound. Kent Beck reaches a similar caution by another route on The Third Loop, where coding models become a 'genie' whose granted wishes rarely match what was meant. The Third Loop opens with four voices, Adam Zimmon, Heidi Waterhouse, Kim Harrison and James Govner, and looks at technology's effect on users through the eyes of the builder, the consumer and, sometimes, the impartial observer. The same episode likens a good release to getting a driver's license: a sense of freedom, with some nerves. Open Source Ready takes the business angle. Its conversation with Sunil Dhaliwal starts from a blunt question: is releasing code about altruism, distribution, community, or reassuring large customers? Then come the knock-on effects, including the risk of competing with a free version of one's own product. AI is framed there as one more platform shift, and like every earlier one it brings new vulnerabilities. Generationship, hosted by Rachel Chalmers, focuses on AI and the infrastructure people have to share. In the episodes here it leans toward the organizational side. Soya Park, introduced as co-founder of a stealth startup, talks about the institutional knowledge that stays in leaders' heads and why best practice is hard to capture when so much of it is implicit. The Dr. Mehdi Nourbakhsh episode describes companies split between excitement and fear about AI, and executives who assume ChatGPT or Copilot is the whole field. There is room for play too, including a riff about an architecture school shaped like Baba Yaga's house on chicken legs. The most technical series is co-hosted by Mark Campbell and Benji DeGroote and talks with maintainers building tools for AI. There, Ivan Burazin explains how sandbox snapshots are spread and layered across nodes, and the episode also carries a stark forecast that sandboxes and RL environments are making CPUs the new constraint. Taken together, the feed covers the developer-tools landscape from server racks to executive workshops.