On Leverage

The most leverage I've found as an engineer, roughly in order:

• Removing work — deleting features, killing projects, saying no
• Improving feedback loops — faster tests, better observability, tighter deploy cycles
• Reducing batch size — smaller PRs, smaller releases, smaller scope
• Building shared understanding — docs, diagrams, ADRs, sync points
• Writing code — surprisingly low on the list once you're past a certain seniority

"The best code is the code you don't write."

A corollary: if you optimise for leverage, you naturally gravitate toward the activities that compound. Code doesn't compound. Knowledge, relationships, and systems do.