The CTO path now runs through shipping code, not managing process

The 2010s management ethos told engineering managers to never make a technical decision again. Will Larson says that era is over. At Stripe he told senior engineers asking for interesting work: "I only have suffering to offer," meaning business-critical problems, not glamorous projects. Back then that looked like bad management; now solving what the business needs is the default path to stability and promotion. The zero-interest-rate era rewarded retention, perks, and headcount growth, and managers became people-pleasers. That world ended.

The CTO mix. Larson describes the CTO role as two capabilities: a strong technologist and a bureaucrat who can run a function, even one they do not deeply know. The old career path through middle management selected for bureaucratic skill and let technical judgment decay. He argues line managers often have better instincts than middle managers because they live in the details. Larson acted on this: he left a 250-person Carta organization to join 25-person Imprint and wrote over 500 pull requests in eighteen months.

The reskill mandate. Middle managers face a hard choice: get back into technical detail or accept that their skills may lose value within a few years. Larson's blunt framing is "am I retiring or reskilling?" Tools for line management and writing software are improving; tools for headcount planning and performance reviews are not. He suggests lateral moves, even a step down, can be educational. The industry's rules changed after a stable decade, and people who came up in that stability feel betrayed.

The org design point. Durable high-ownership teams are the only way to scale, but they look wrong optically. The alternative is running people to the latest fire, which works until there is a crisis everywhere. At Stripe he shifted from a "defend the castle" model to pods that own areas, and stopped managing them until something broke. Slack shows this: its agents integration shipped because the owning team had domain context while competitors did not respond. Domain knowledge compounds when teams stay intact.

The ladder caveat. Career ladders still matter for fairness, but detailed rubrics get gamed. At Carta, cost pressure led him to impose a hard staff-engineer ratio, a constraint that forced more junior hiring and a real pipeline. At Imprint, the goal is velocity, so the ratio would not make sense. The staff-plus signal is still decision-making in ambiguous situations, not raw coding. Agents can make code easier, but they do not carry the implicit context and tradeoffs that live outside the context window.