5 Things to Know About Context Engineering

Software systems are getting better at understanding themselves. The mix of richer telemetry, smarter pipelines, and agentic AI is shifting observability from a passive record of events into something more active and useful. That shift is what we mean by context engineering.

We recently partnered with O’Reilly on a report by David Beale that introduces the discipline. Before you read it, here are five things worth knowing.

1. Two problems are converging.
   SRE and platform teams are buried under telemetry they cannot fully reason about. At the same time, the AI agents they deploy to help are inheriting that same confusion. Both issues come from the same design mistake: treating telemetry as a dump of events instead of a structured source of truth.

2. AI does not fix bad telemetry. It amplifies it.
   A mislabeled log or a missing trace can flow through embeddings, vector stores, and reasoning layers and still produce a confident answer that is wrong. The report makes a strong case that better AI outcomes in operations start with better context, especially structured signals that carry service lineage, dependencies, and intent.

3. Active telemetry is different from storing everything.
   Instead of emitting the same data regardless of system state, active telemetry adapts. Signals can describe themselves, enrich or compress based on the consumer, and connect cause and effect in ways that guide both human and AI decision-making. That changes the economics of observability.

4. Context has to be designed in from the start.
   This is where context engineering becomes practical. It means designing data, metadata, and feedback loops so both humans and machines can reason about what is happening. That includes context schemas, enrichment pipelines at the edge, and distributed context graphs that evolve with the systems they describe.

5. The SRE role is shifting.
   As systems become better at self-description and bounded automation, the highest-value work for SREs moves away from reactive firefighting and toward building the feedback systems that power human and machine reasoning. The report explores what that looks like in practice, including how to design for autonomy and how to build trust for machines acting inside clear limits.

The full report goes deeper on the architecture, with diagrams, design principles, and a practical way to get started.  Read "Context Engineering for Observability" at mezmo.com/oreilly.

‍