SED News: OpenCode, AI Code vs. Shipped Code, and the LiteLLM Breach

In this episode of SED News, hosts Gregor Vand and Sean Faulconer dive into the latest tech headlines, focusing on the resurgence of CPUs amid the rise of agentic AI workloads, the LiteLLM security breach linked to a compromised SOC 2 audit, and the growing competition between open-source and commercial AI coding tools. The central theme explores the critical gap between writing code and shipping it: while AI accelerates code generation, the validation and review layers—especially human-in-the-loop processes—have not kept pace, leading to a bottleneck in production deployment. Data from CircleCI’s 2026 State of Software Delivery report reveals that while top-performing teams doubled their throughput, the median team saw only a 4% increase, and main branch throughput actually declined by 7% despite a 50% surge in feature branch creation. This highlights a growing crisis in software delivery where speed of generation outstrips the capacity for quality control, risk management, and security review. The hosts warn that overemphasizing rapid prototyping over productionization risks more outages, security flaws, and cultural misalignment in engineering teams. They also reflect on how AI is democratizing software ideation, enabling non-technical users to prototype ideas quickly, but caution that this shift demands new organizational structures and a reevaluation of where human oversight is most critical. The episode closes with a lighthearted look at Hacker News favorites, including Doom running over DNS, the psychology behind seafoam green control room walls, and a Tesla Model 3 computer built from crash debris.