A message from our sponsor, Archera:
AI workloads are unpredictable, which makes cloud commitments feel like a gamble. Archera insures your commitments against underutilization, so you can push coverage higher without the risk of getting stuck. If usage drops, Archera covers the downside. Commitment Release Guarantee included.
Improving robustness requires increasing complexity. Let’s throw more complexity at it?
I’m using this enormously complex system, an LLM, to help me solve a problem that was created by software complexity in the first place.
Lorin Hochstein
This feels like using multiple agents as a sort of redundancy and cross-validation architecture to improve the reliability of agent output..
Alex Ewerlöf
This article explains why end-to-end testing breaks down in microservice-based systems, not due to poor tooling, but because of fundamental architectural and operational mismatches.
Alok Kumar — DZone
LaunchDarkly’s survey data show have some interesting things to say about the impact of AI.
[…] while build and deployment velocity have improved, production reliability has not.
LaunchDarkly
Fred Hebert surveyed how AI coding assistants vs. AI SRE tools are marketed and found a stark divide: coding assistants are framed as partners that augment engineers, while AI SREs are framed as replacements for low-value work. The implication is that the people building and buying these tools see incident response as grunt work to be automated away — and that says a lot about how decision-makers perceive the role.
Fred Hebert
I especially like the point that incidents are leadership moments — how you respond tells your team everything about the culture you’re building. This one is aimed at CTOs, but really it’s a great reminder for anyone in a leadership role during incidents.
Joe Mckevitt — Uptime Labs
There’s a really interesting bit in this one about libraries and layers of the system doing their own retries without your knowledge, magnifying retry volume.
David Iyanu Jonathan — DZone
I like the section on what AI should and shouldn’t do. It’s important to avoid automating away the process of learning from incidents.
incident.io