Software Engineering Podcasts

Python Bytes, with Michael Kennedy and Brian Okken, works at the scale of one language. In #474: Astral to join OpenAI (46 minutes), the hosts weigh why the Astral team might join OpenAI, and one flags his reasoning as speculation. A host's complaint that Claude Code keeps searching through files instead of understanding the project turns the news into a lesson about how coding tools read code. On The Changelog: Software Development, Open Source, The GitHub problem (and other predictions) (Friends) (101 minutes) is looser. Riffs about Chandler Bing and a childhood tennis detention threat sit alongside an argument that GitHub "earned its monopoly" and then stagnated. It rewards listeners with patience for tangents.

The Stack Overflow Podcast covers similar ground in a shorter episode. In Prevent agentic identity theft (25 minutes), Nancy Wang, CTO at 1Password, compares public and private keys to a bank vault. Asked about brute-force attacks, she answers, "I'm not a crypto expert here," and falls back on the company's own account of its breach history. A careful listener will hear both the candour and the claim.

On The Changelog: Software Development, Open Source, Selling SDKs in the era of many Claudes (Interview) (110 minutes) is about business. Its guest explains trial licence keys, pricing calls and why an SDK is "built entirely in anticipation of third party code." On Talk Python To Me, OWASP Top 10 (2025 List) for Python Devs (66 minutes) stops on a popular CDN's suggested embed code and asks what is missing: an integrity check. That moment is partly visual, so listen near a screen. For more on the security side, see the Cybersecurity guide.

The stronger AI episodes attach their opinions to evidence. On PurePerformance, AI-Ready Codebases: Engineering Discipline for Agentic AI with Adam Tornhill (52 minutes) brings numbers rather than mood. Tornhill describes a code-health score built from 25 "code smells," which are signs of poorly structured code, scored out of 10. He says the industry average is 5.15. He also argues that his research shows a codebase needs at least 9.5 before agents can speed up work while keeping AI-induced defects low. Those are his findings, offered as his findings, and they give a sceptic something to test.

Arrested DevOps takes the human angle. In How AI Is Changing the SDLC With Hannah Foxwell and Robert Werner (40 minutes), one speaker argues that AI "has taken the actual typing off you, but it's not taken the thinking." The conversation then turns to the pressure on developers whose leaders expect faster shipping. On Software Defined Interviews, AI Doesn’t Fix Bad DevOps: Lessons from 15 Years of DORA Data, with Nathen Harvey (75 minutes), the guest explains that his team uses AI to vet survey questions, but humans write the reports.

The forecasts get bolder elsewhere. On Telugu Bytes, 087 - Tsun"AI" Warning 🌊 (104 minutes), the hosts compare the moment to Wuhan in December 2019. They state their recording date, February 6th, 2026, so the warning can be checked later, and they guess that early-career hiring will contract sharply. Treat it as a prediction, delivered with conviction. On a Talk Python To Me panel, 2025 Python Year in Review, one voice suggests an AI bubble could crash the economy while agentic coding tools survive. The AI & ML Engineering page and the Artificial Intelligence guide go deeper.

Some shows ask what good work feels like. On Code with Jason, 318 - Adam Dawkins, CTO of Dragon Drop (53 minutes), Jason asks, "What is good code?" Dawkins answers readability and editability. He warns against a "CSSification of objects," where developers keep stacking new code on top because they do not know where to start with the old. Jason reports that someone else answered more briefly: easy to change.

Porozmawiajmy o IT, in Polish, approaches failure calmly. In Systemy IT się psują. Co z tego wynika i czego nas to uczy. Gość: Adam Korga - POIT 313 (49 minutes), Korga argues for postmortems that name what happened without hunting for blame, and mentions his own DROP DATABASE command run in the wrong terminal. Developer Tea compresses its lessons into short sessions. Your Capacity for Growth Is Dependent on This Factor - Cognitive Load Theory (24 minutes) borrows from flight training, where instructors push students to the edge of task saturation.

On Mainframe Connect, Sudharsana Srinivasan advises newcomers to join a user group where they can request a mentor. The Engineering Leadership page gathers more of this. For listening from Python Bytes to .NET Rocks! to Syntax, see the Languages & Craft page.

Operations shows sit closer to running systems. PerfBytes' Talking about Kubernetes (56 minutes) explains vertical and horizontal scaling as adding pancakes to one stack versus setting out another plate. The DevOps & Cloud page continues from there, and the Technology guide maps the wider section.