AI is making it dramatically easier to create and change software, but writing code is only one part of the software lifecycle. Every change still needs to be deployed, observed, operated, diagnosed, and kept reliable in production. As the volume of software changes grows, the bottleneck increasingly shifts from writing code to operating it safely.

Drawing on lessons from operating large-scale distributed systems at LinkedIn, this talk follows a software change through the production lifecycle: Deploy → Observe → Operate → Incident → Diagnose → Mitigate → Learn → Improve. We will look at practical patterns for catching problems before production, reducing deployment risk, detecting and diagnosing failures, and deciding where production responsibilities should live as more people and AI agents create software.

Finally, we will look at how incidents can become inputs to the next software change by turning operational learnings into automated tests, deployment policies, detection rules, rollback conditions, and platform guardrails.

The goal isn't to slow down software development. It's to make sure we can move faster without sacrificing reliability, with platforms that make production safety the default.

Key Takeaways:

• Learn how to evolve the SDLC beyond AI-assisted coding, with practical patterns for safely deploying, observing, diagnosing, and mitigating a growing volume of software changes.
• Understand which production responsibilities belong in the platform versus the application, and how abstractions and guardrails around deployment, observability, traffic, reliability, and rollback can make safe operation the default.
• Learn how to build a feedback loop from production back into development, turning incidents and operational learnings into automated tests, detection rules, deployment policies, and guardrails that prevent the same class of failure from recurring.