Will Larson on Engineering Strategy, Systems Thinking, and Writing Consistently
Will Larson is CTO at Carta, formerly Stripe, Uber, and Calm, and the author of An Elegant Puzzle, Staff Engineer, and The Engineering Executive’s Primer (O’Reilly). This episode covers how to think about engineering strategy (why every company already has one and why writing it down is the only way to improve it), systems thinking applied to engineering organisations, the EM–PM relationship, how to measure engineering productivity honestly, and how Larson has written more than a thousand posts across sixteen years without burning out.
Key ideas
- Every company has a strategy; most are just unwritten. The complaint “we have no engineering strategy” is always wrong — there is always a strategy, sometimes bad, always implicit. The moment you write it down you can improve it. Unwritten strategy is impossible to debug: you cannot tell whether a PM is applying it incorrectly, whether it is poorly designed, or whether it is simply not being communicated. Writing makes it legible.
- Boring strategies are powerful. Good engineering strategy is usually anti-exciting. Examples: Carta’s “standard kit” (only the tools we already use), Uber’s “no cloud, own data centers,” Stripe’s “Ruby monolith.” Each constrained engineers in ways that frustrated them while focusing energy on problems the company actually cared about. The goal of strategy is not to make engineers happy; it is to allocate limited capabilities toward the problems that matter.
- Rumelt’s three-part definition. Diagnosis (what is true about the current situation?) → Guiding policies (given the diagnosis, how do we want to address the constraints?) → Actions (how do we implement the guiding policies?). Bad strategies almost always trace to a bad diagnosis — specifically, a willful disbelief about real constraints (“we can do all of these projects at once”).
- Systems thinking: stocks and flows. Stocks are things that accumulate (engineers, open roles, code quality). Flows are rates of change between stocks (hiring rate, attrition rate, technical debt accumulation). The hiring pipeline modelled as a stock/flow system reveals why cutting headcount budget never solves a velocity problem that is actually a retention problem.
- Measuring engineering productivity honestly. DORA metrics (deployment frequency, lead time, change failure rate, time to restore) are useful diagnostic tools but poor evaluation metrics. They tell you where to investigate, not whether to fire engineers. The more honest signal: talk directly to engineers. They know whether their teams are effective and, critically, why not.
- Writing consistently. The secret is writing only what gives you energy, never on assignment, never chasing what is popular. Larson wrote roughly a thousand posts in sixteen years while holding full-time engineering leadership roles at hypergrowth companies. He attributes the sustained volume to writing about what he is thinking about right now, not what he thinks people want. The biggest risk to writers is quitting; writing what does not energise you makes quitting more likely.
- Shared performance ratings for EM–PM pairs. At Carta, engineering managers and PMs calibrate on the same performance rubric with their scores set together. This structural alignment removes the common tension where an EM protects the team’s stability while a PM advocates for a packed roadmap — both are evaluated against shared outcomes.