The Stack Overflow Podcast Revealing the unknown unknowns in your software
18 snips
Nov 7, 2025 Nic Benders, Chief Technical Strategist at New Relic, dives into the intricacies of modern software systems. He discusses the complexity crisis and the need for engineers to understand lower layers of abstraction. Benders emphasizes the difference between observability and understandability, highlighting AI's dual role in complicating and aiding system management. He advocates for open telemetry for better AI visibility, and stresses the importance of demystifying AI to enhance trust and control in software development.
AI Snips
Chapters
Transcript
Episode notes
Complexity Grows With Each Abstraction
- Software complexity has exploded as we moved to microservices, cloud, containers and orchestration, widening the stack above the hardware layers.
- Each abstraction added capability but also removed visibility, making debugging like solving a distributed murder mystery.
Observability Isn’t Enough
- Observability gives you visibility and historical context, but it alone doesn't produce understanding of complex systems.
- Nic argues we need 'understandability' on top of observability to reason about and act on system behavior.
Log AI Interactions And Analyze Them
- Monitor AI interactions by recording inputs, outputs, latency and cost to allow later inspection and anomaly detection.
- Use AI to analyze responses (e.g., sentiment) and flag problematic interactions for review.