The Confident Commit

Spinning up an agent is like hiring a brilliant engineer on day one: capable of almost anything, completely in the dark about how your organization actually works. That gap is expensive, and most teams are only starting to feel it.Dennis Pilarinos, founder and CEO of Unblocked, has been thinking about this problem longer than most. His argument is that the hardest part of software engineering was never writing code. It was always the context layered around it: the architectural call buried in a Slack thread, the incident postmortem that quietly changed a naming convention, the tribal knowledge that lived in someone's head until they left. AI made that problem viscerally visible. Agents burn tokens chasing answers that experienced engineers already carry.Dennis and Rob dig into what it actually takes to compete on context beyond just indexing source code, why the pull request is quietly becoming the uber class of the modern SDLC, and why the tokenomics reckoning that's coming for engineering leaders might finally give us the through-line from AI spend to business outcomes that the industry has never had.Dennis built developer tools across every major platform shift of the last 20 years, from the .NET framework at Microsoft to AWS, mobile CI at Buddybuild, and now agentic infrastructure at Unblocked.Why agents are "disgustingly inefficient" without organizational contextThe limits of source code alone, and what fills the gapHow the pull request is changing form and function in an agentic worldContext-powered code review and what it surfaces that standard tools missConnecting AI spend to business outcomes: closer than ever, further than we thinkThe 3-5% of engineers who are "AI-pilled" and what they've already figured outSubscribe wherever you get your podcasts.

Everyone keeps saying we don't need junior engineers anymore. This episode makes the case that's exactly wrong.Rob sits down with two CircleCI colleagues at opposite ends of the experience spectrum: Hanabel Mengistu, a new grad less than a year into her first engineering role, and Michael Webster, who has been writing software professionally since 2012. Together they explore what it actually feels like to enter the industry right now, when AI has reshuffled the deck for everyone.The conversation covers the gap between what college teaches (computer science) and what the job actually requires (everything else), the apprenticeship model and how AI is quietly eroding it, and why Hanabel's lack of preconceptions turns out to be a genuine advantage. Rob frames it well: when the tools change fast enough, everyone is back at the beginning together. That's disorienting for veterans, but it might be the best possible moment to be starting out.Hanabel Mengistu is part of CircleCI's Associate Rotational Program. Webster is a Staff Engineer at CircleCI with over a decade in the industry.Topics covered:Why "we don't need junior devs" is a short-sighted takeThe college-to-industry gap: CS education vs. real software engineeringHow AI is disrupting the apprenticeship model (and creating isolation for juniors)The "great leveler" effect: when nobody has 10 years of experience in tools that are 1 year oldCuriosity as the defining trait for engineers in 2026Essential vs. accidental complexity (and why Fred Brooks still matters)Subscribe wherever you get your podcasts.

Code review has always been a time sink. AI just makes the dysfunction undeniable.Luca Rossi, founder of Refactoring.fm and builder of the open source tool Tolaria, has been running one of engineering's most-read newsletters for five years, with over 170,000 subscribers. He's also been doing what a lot of engineering leaders talk about but rarely do: building a real product with AI agents to pressure-test what's actually possible today.In this conversation, Rob and Luca dig into the state of AI adoption at the midpoint of 2026; what high-performing teams are getting right, why most orgs are still just bolting AI onto a broken process, and why code review was already a questionable practice before AI came along. Luca also walks through the "guides, gates, and guards" framework he uses to keep AI agents productive and honest on Tolaria: the instructions that steer agent behavior, the deterministic hooks that catch what agents ignore, and the nightly reflection loops that catch the rest.Luca is the founder of Refactoring.fm, a newsletter and community covering software engineering and engineering leadership. He spent a decade as a CTO and startup founder before going independent, and now builds in public as both a writer and a solo product builder.In this episode:Why adopting AI without fixing your process just means moving faster in the wrong directionWhat the most mature engineering teams are doing differently in 2026The guides, gates, and guards framework for AI-assisted developmentWhy Luca has been a code review skeptic since before the AI eraHow building Tolaria became a way to get ground truth about what AI agents can actually doThe unexpected differences between running a startup and running an independent content businessSubscribe wherever you get your podcasts.

Most AI applications in production right now were shipped on vibes. A developer ran a few queries, liked what they saw, and pushed to prod. Laurie Voss argues that's the core reason so many AI products feel broken, and the fix is simpler to name than it is to do: write the tests.Laurie is Head of Developer Relations at Arize AI. He's spent his career watching developer behavior at scale, and now building the case for rigorous eval practices in AI engineering. In this episode, he makes the case that evals aren't a specialized ML concept but a natural extension of software testing discipline that most engineers already understand.Rob and Laurie dig into what makes evals hard (non-determinism, the need for LLM-as-judge), how to keep costs manageable without sacrificing coverage, and the emerging pattern of capability evals that drive software improvement without any human in the loop. They also get into context engineering, context graphs, and whether software development is a narrow enough domain for agents to one-shot it anytime soon.Topics covered:• Why "evals are just tests" is the reframe that unlocks adoption• LLM-as-judge: how it works and how to tune it• Regression evals vs. capability evals, and why the distinction matters• How to run evals cost-effectively without going to Opus for everything• Context graphs as a compression strategy for large domains• Whether non-technical builders can ship reliable software with agents todaySubscribe wherever you get your podcasts!

If 90% of code is now AI-generated, what happens to the engineers who never had to write it themselves? The ability to review code well, to catch what the LLM missed, depends on mental models that only come from practice. That tension sits at the center of this conversation.Rob and Hywel dig into how software engineers actually develop expertise, why knowledge transfer and real learning are not the same thing, and what it means to grow junior engineers in an era when the reps that used to build skill are disappearing. They also get into what separates engineers who get serious leverage from AI tools from those who don't, and why the answer has less to do with prompting tricks than with understanding how LLMs actually work.Hywel Carver is the founder and CEO of Skiller Whale, a human-led technical learning platform focused on building engineering skills through coached, hands-on practice. His teams have worked with organizations including Pleo, where one cohort saw a 235% productivity improvement through structured skills development.Topics covered:Why knowledge transfer is not the same as learningBloom's taxonomy and what it takes to evaluate code effectivelyHow AI adoption is changing the skill development path for junior engineersWhat separates high-leverage AI users from low-leverage onesMental models as the durable foundation for navigating constant changeWhether agile principles still hold in an AI-accelerated worldSubscribe wherever you get your podcasts.

Most teams obsessing over token costs are measuring the wrong thing. The real savings from AI aren't in lines of code written faster. They're in the coordination overhead that disappears when fewer humans need to align before anything gets built.Chris Kelly, Head of Product at Augment Code, joins Rob to cover why prototypes have replaced specs, how agents enable dynamic team capacity the way cloud replaced over-provisioned servers, and what "good code" even means when your primary reader is an LLM.In this episode:Why coordination costs are the hidden savings of AI-assisted developmentHow agents make dynamic team capacity finally achievableWhat junior engineers actually need to know as abstraction layers keep risingHow self-improving agents could change incident response and code qualityHave someone you’d like to hear on the show? Reach out to us on X at @CircleCI!

What does it actually look like to build an AI-native product and lead an engineering team through the AI era when you've been doing it longer than most? Rob Zuber sits down with Loïc Houssier, CTO at Superhuman, to talk about what it meant to be an AI company before AI was everywhere, and how that early foundation shapes the way they build, ship, and think today.The conversation covers how Loïc drove AI tool adoption across his engineering org without mandates (and which senior engineer's change of heart became a cultural turning point), why great UX is still the real moat in an age where anyone can ship an average product fast, and how email, despite everything, remains the connective tissue of professional life. Plus: what it's like to rethink your entire SDLC when the economics of building software change overnight.Have someone you’d like to hear on the show, reach out to us on X at @CircleCI!

In this episode of The Confident Commit, Rob Zuber sits down with Greg Foster, Co-Founder and CTO of Graphite, for a timely conversation about what it really means to ship software fast in the age of AI. Greg challenges the "996" grind culture with a provocative billboard campaign, arguing that speed comes from smart tooling and agency rather than endless hours, and shares how AI code generation is fundamentally reshaping the code review bottleneck while creating new trust and context challenges for engineering teams.The conversation explores the paradox of our current moment—where developers can generate more code than ever, yet senior engineers wake up to 20 AI-generated PRs that still require careful human review. Greg discusses why companies like Google’s techniques for handling high volumes of code changes are suddenly relevant again, the emerging role of AI code review as a "super linter," and why he believes the best engineering cultures prioritize being "urgent, expert, and kind." Whether you're an engineering leader navigating the AI transformation or a developer trying to understand how code review needs to evolve, this episode delivers practical insights on maintaining quality and velocity as AI reshapes software delivery.Have someone you'd like to hear on the show? Reach out and let us know on X at @CircleCI!

In this episode of The Confident Commit, Rob Zuber sits down with Johnny Austin, VP of Engineering at Upwell, for an honest conversation about what it really takes to grow as an engineering leader. The conversation explores the critical distinction between making your team happy versus setting them up for long-term success, why the most impactful engineers are those who ask questions about the business and product rather than just technical implementation details, and how to effectively communicate context across multiple levels of an organization. Whether you're an early-career engineer learning to think beyond the code or a leader working to build teams that drive real business value, this episode delivers practical wisdom grounded in real-world challenges.Have someone you'd like to hear on the show? Reach out and let us know on X at @CircleCI!

In this episode of The Confident Commit, Rob Zuber sits down with Meg Adams, Senior Director of Engineering at The New York Times, for a deep dive into leading engineering teams through the AI revolution while staying true to organizational mission. Meg shares how the Times approaches AI adoption with a "measured but focused" strategy, emphasizing experimentation and opinion-formation over mandates, and why she believes AI serves as a force multiplier for what already exists in your organization and workflows.The conversation explores the unique challenges of product engineering in media, from building cooking products for users across the entire spectrum of culinary expertise to navigating digital transformations that honor legacy while embracing an uncertain future. Meg discusses the importance of mission-driven leadership through technological transitions, why she prioritizes helping her engineers stay valuable in the marketplace, and her unconventional path from elementary education to retail management to engineering leadership. Whether you're an engineering leader managing AI adoption or someone thinking about the intersection of technology and human-centered products, this episode delivers insights on building resilient teams that can adapt while maintaining their core purpose.Have someone you'd like to hear on the show? Reach out and let us know on X at @CircleCI!