Engineering leadership · Book notes
If you keep one thing from The Staff Engineer's Path, keep the maps
Now that execution is cheap and verification is the real bottleneck, the locator and topographic maps are the terrain only a senior engineer can read.
The Staff Engineer’s Path by Tanya Reilly (O’Reilly, 2022) has three parts. The single strongest candidate for a takeaway is the set of three maps in part one (the big picture). “At the staff engineer level, everything you do has a high opportunity cost, so your work needs to be important.” Senior individual contributors and architects in large enterprises are judged on this question, and the maps are the tool this book gives you to answer it.
Map 1 (the locator map, or “you are here”): Understand your team’s purpose and how your team’s output affects the teams that depend on it. Map 2 (the topographic map): Understand the terrain between teams.. the friction points, where decisions are really made, the informal org chart that exists beyond the formal org chart. Reilly says you can learn this map by using open sources like shared calendars and chat channels. Map 3 (the treasure map, or “x marks the spot”): Understand where everyone wants to get to and why it’s important to them.
Part two, execution: This is where the picture has changed the most since 2022. Maxime Najim in LeadDev talks about how for staff-plus engineers, verification has replaced execution as the bottleneck. A staff engineer can bring a working prototype or a ready-to-go pull request instead of a ticket in the backlog. The treasure map is much easier: an agent can build an “x marks the spot” demo in an afternoon, and people can react to something real.
But the locator and topographic maps aren’t easier. The 2025 DORA report talks about how AI amplifies whatever system it enters, and the payoff from platform quality and team alignment is the terrain of those two maps. An agent can read your repository, but it can’t see the informal org chart, the risk committee that owns a decision, or which shared platform several business units depend on. For large regulated organizations like banks, that terrain is most of the work.
The trust numbers tell the same story: DORA found that about 30 percent of respondents had little or no trust in AI-generated code. So somebody needs to know where that output lands and who it affects. I’ve spent the last ten years working with large financial institutions and Fortune 500 companies, now as a principal cloud architect, and before that I was a developer and a sysadmin. Every role showed me a different slice of the same map: how code connects to the people and systems around it.
The idea strains in places. The book has been argued about for a long time, and some of that’s about how staff titles have been diluted, how it reads like a promotion guide for big tech, how somebody still has to fix the bugs. Part three, on leveling up, might be the hardest to apply right now. Stanford’s Digital Economy Lab has documented declines in early-career employment in AI-exposed jobs like software development, while experienced workers have stayed steady. That means fewer people are learning the terrain slowly. If maps are a scarce senior skill, who teaches the next generation how to draw them when the junior work that taught them goes to agents?