Overcommitted | Software Engineering and Programming Insights

What happens when coding becomes nearly free - but everything else in the software delivery process doesn't? Cameron Etezadi, CTO at LaunchDarkly, joins Erika, Brittany, and Bethany to talk about the gap between code that compiles and code that actually works in production. They cover vibe coding pitfalls, the release-observe-iterate loop, how AI is shifting engineers toward frontline manager behaviors, and why speed without control is a liability - not an asset. Cameron also shares how LaunchDarkly changed their intern interview process to reflect the AI era, what EU regulations mean for human accountability in AI-generated code, and why the classic two-pizza team is giving way to something much smaller.
Cameron Etezadi is the CTO at LaunchDarkly and brings experience scaling engineering organizations at Google, HashiCorp, and SAP. He's also a licensed flight instructor with ratings in fixed-wing, seaplane, helicopter, and gyroplane - and has logged over 400 skydives.

Show notes:
Cameron on LinkedIn: https://www.linkedin.com/in/cameronetezadi
LaunchDarkly: https://launchdarkly.com
LaunchDarkly leadership page: https://launchdarkly.com/leadership/
Rainier Flight Service bio: https://www.rainierflightservice.com/rainier-flight-staff/cameron-etezadi

Summary
Brittany's first weeks at Bluesky are reframing everything she thought about coding interviews. Career growth in tech still hinges on algorithm problems, but AI, remote work, and async development have fundamentally changed what actually signals engineering ability. Why do companies cling to LeetCode when software engineers now solve problems in decentralized, AI-augmented teams? We dig into what hiring should measure, why job-hunting feels unsettled, and how women engineers navigate visibility and advancement in the interview process.

Links
How GitHub Engineers Learn New Codebases blog post: https://github.blog/developer-skills/application-development/how-github-engineers-learn-new-codebases/
Diataxis: https://diataxis.fr

Hosts
Overcommitted: https://overcommitted.dev
Bethany Janos: https://www.trustyduck.dev/
Brittany Ellich: https://brittanyellich.com
Erika Eggemeyer: https://github.com/eggyhead

Summary
Dennis Pilarinos has spent nearly two decades building developer tools at Microsoft Azure, AWS, and Buddybuild (acquired by Apple). Now CEO of Unblocked, he's tackling the critical problem holding back agentic AI development: context intelligence. Why do AI coding agents keep guessing? Because they lack the organizational and business context behind the code. In this episode, the Overcommitted hosts dig into how developers and AI agents can truly understand the "why" behind codebases, not just the "what"—and why this matters for code quality, productivity, and shipping reliable tools.

Links
Dennis' email: dennis@getunblocked.com
Unblocked: https://dub.sh/IeI82Vx
Dennis on X: https://x.com/dennispilarinos
Dennis on LinkedIn: https://www.linkedin.com/in/dennispi/

Hosts
Overcommitted: https://overcommitted.dev
Bethany Janos: https://github.com/bethanyj28
Brittany Ellich: https://brittanyellich.com
Erika Eggemeyer: https://github.com/eggyhead

Summary
What does it actually look like to hand 90% of your workday over to an AI agent you built yourself?
Sterling Chin, founding DevRel at Inngest and creator of Marvin, an open-source AI chief of staff with nearly 1,000 GitHub stars, has been doing exactly that for months. In this episode, Sterling joins Brittany to talk through how Marvin works, why he built it, and what he's learned about the real friction points in AI adoption that most people don't talk about.
Sterling came into tech through a coding bootcamp after studying elementary education at BYU, landed at Postman where he led the R&D labs team, went viral on LinkedIn for posts about their AI assistant PostBot, and accidentally became a DevRel engineer because of it. Now at Inngest, he's the founding DevRel hire, and still building in public constantly.

Links
Sterling's Website: https://sterlingchin.com/
Sterling's LinkedIn: https://www.linkedin.com/in/sterlingchin/
Sterling on GitHub: https://github.com/SterlingChin
Sterling on Bluesky: https://bsky.app/profile/sterlingchin.bsky.social
Sterling on Twitter: https://x.com/SilverJaw82
Marvin: https://github.com/SterlingChin/marvin-template
Sterling on YouTube: https://www.youtube.com/@SterlingChin
Sterling on Substack: https://sterlingchin.substack.com/
Web Dev Challenge Episode: https://www.youtube.com/watch?v=X2sEoZG8EIw&list=PLz8Iz-Fnk_eTkZvSNWXW_TKZ2UwVirT2M&index=17
Inngest: https://www.inngest.com/
AI Crimes in Production: https://ai-crimes-in-production.com/

Hosts
Overcommitted: https://overcommitted.dev
Brittany Ellich: https://brittanyellich.com

AI is generating more code than ever, but most engineers aren't verifying it. Sonar Staff AI Researcher Joe Tyler shares breakthrough findings from his LLM Leaderboard research on code quality, the hidden "coding personalities" of different models, and why the real bottleneck in software engineering isn't writing code: it's securing and reviewing it. Discover the gap between developer distrust and actual verification practices, plus how to position yourself for the verification-first future of software development.

Topics: AI code quality, LLM research, software engineering careers, code verification, developer tools

Show links:
Sonar LLM Leaderboard: https://www.sonarsource.com/the-coding-personalities-of-leading-llms/leaderboard/
Sonarqube: https://www.sonarsource.com/products/sonarqube
Sonarsweep: https://www.sonarsource.com/products/sonarsweep/
Joe's LinkedIn: https://www.linkedin.com/in/joe-tyler-a668051b1/
Latent Space: https://www.latent.space/
Turing Post: https://www.turingpost.com/
Nathan Lambert: https://substack.com/@natolambert
Cameron Wolfe: https://substack.com/@cwolferesearch
Sebastian Raschka: https://substack.com/@rasbt
Andrew Ng: https://www.andrewng.org/