Great, rapid, reliable engineering is rarely about single breakthroughs. It is about designing systems where ordinary work compounds quietly, predictably, and in your favor, increasingly accelerated by agentic workflows.
Early in my career, I chased leverage through big changes: rewrites, new architectures, breakthrough tools. Those moments exist, yet they are rare and expensive. What consistently moves systems forward is smaller: a sequence of correct decisions that make the next decision cheaper, faster, and safer.
That became clear once we started treating the engineering process itself as something worth engineering. Instead of focusing only on what code did, we focused on how it was written, reviewed, and changed.
One example is how we work with agents. We treat them as engineers who follow instructions literally, so we assign narrow responsibilities and make those responsibilities explicit in the repository. When an agent opens a pull request, it may update AGENTS.md alongside the code. If a failure mode appears twice, the fix becomes a rule. These can be simple. For example, we added a rule to never use bare except blocks and updated SKILLS so pull request titles explain why a change was made, which helps maintain consistency across development.
This is compound engineering. Corrections become guardrails. The system shifts from relying on vigilance to relying on structure. Breakthroughs matter, yet they are episodic. Compound engineering is continuous, turning judgment into infrastructure so ordinary work compounds quietly over time.
Subscribe for new essays
Get updates when new writing goes live.
