Ep. #44, Beyond the Model: Observing AI Agents with Mikyo King

On episode 44 of Open Source Ready, Brian Douglas and John McBride sit down with Mikyo King to explore why an agent can fail even when its model gets things right. They discuss observing the full agent workflow, using evaluations to uncover unexpected behavior, and testing whether skills actually improve results. Along the way, Mikyo shares how Arize approaches open source collaboration as coding agents reshape the contribution process.

On episode 12 of Third Loop, the Progressive Delivery team speaks with Steph Hippo, Platform Engineering Director at Honeycomb, about what happens when AI accelerates development beyond the pace humans can absorb. They explore safe defaults, evolving code reviews, and the feedback loops that help teams understand what they ship. Steph makes the case for reinvesting saved time in better decisions, software quality, and human connection.

On episode 59 of Generationship, Rachel Chalmers speaks with Emily Long and Alex Zenla, co-founders of Edera, about rethinking the foundations of cloud and AI security. They explain how an IoT side project became an approach to workload isolation built around Xen and Rust, and why AI agents are making stronger security boundaries increasingly urgent. Along the way, they discuss complementary leadership skills, fundraising lessons, and the value of questioning the industry’s technical assumptions.

In episode 5 of Lab Notes, Amir Zohrenejad speaks with Aoden Teo, co-founder and CEO of Miso Labs, about what it will take for voice AI to become truly humanlike. They explore the limitations of today's speech-to-text and text-to-speech pipelines, the promise of full-duplex speech models, and the technical challenges around audio tokenization, turn-taking, emotion, reasoning, and real-time inference.

On episode 93 of o11ycast, Ken Rimple and Martin Thwaites sit down with Mike Goldsmith to explore how teams can control telemetry costs without sacrificing the data they need to understand their systems. They discuss choosing the right signals, reducing repetitive logs, building more efficient OpenTelemetry Collector pipelines, and using dynamic sampling to preserve valuable traces while cutting unnecessary volume.

On episode 11 of Third Loop, the Progressive Delivery team sits down with Chad Fowler to explore regenerative software and what architecture looks like when AI makes code cheap enough to continually replace. They discuss Phoenix Architecture, Pace Layers, observability, architectural constraints, and why the system itself, rather than any individual implementation, may become the real asset in an AI-native future.

On episode 58 of Generationship, Rachel Chalmers sits down with Anastasia Marchenkova. They explore what it will take to move quantum computing from promising hardware to useful, production-ready systems, including better orchestration across quantum, classical, and AI compute. Anastasia also discusses open-source infrastructure, the limits of AI automation, and her larger vision of making powerful computing and scientific tools more broadly accessible.

On episode 4 of Lab Notes, Amir Zohrenejad speaks with Junchen Jiang about why KVCache may be better understood as reusable, AI-native data rather than a temporary inference optimization. They explore how LMCache and CacheBlend can reduce redundant computation, move context across distributed inference systems, and help support increasingly complex AI agents. The conversation also covers multimodal workloads, open-source infrastructure, and the future of AI systems research.

On episode 10 of Third Loop, the Progressive Delivery team speaks with Honeycomb co-founder and CTO Charity Majors about observability, AI, and the changing economics of software development. They explore a future where code is increasingly disposable and regenerable while architecture, constraints, production behavior, and promises to users become the durable artifacts that matter. Along the way, Charity shares her perspective on responsible AI adoption, engineering in nondeterministic systems, and why production remains the ultimate source of truth.

On episode 12 of High Leverage, Joe Ruscio sits down with Dexter Horthy of HumanLayer. They explore the promise and limitations of autonomous coding, including why today’s models excel at bounded programming tasks but struggle to account for the long-term consequences of architectural decisions. The conversation covers dark software factories, code review bottlenecks, program design, technical debt, and the continuing importance of experienced engineers.