The revised rules of engineering management with Will Larson
Will Larson argues that zero-interest-rate habits no longer serve engineering management: leaders must return to technical details, engineers must solve business problems, and line managers make better executives than middle managers.
Watch on YouTube ↗Charlie's summary
- ▶ 19:28
Dig v4 died of a launch with no rollback, an untested Cassandra, and a mutable-default-argument bug in Python
Will joined Dig three months before its rewrite to Cassandra. Load testing was handed to two engineers two days before launch, and they found reads and writes could not run together. The old servers were wiped, so rollback was impossible, and the site stayed down about 24 hours while a launch party went on. Memory leaks forced hourly restarts until Will spent four weeks tracing the cause to a Python list used as a default parameter, which kept every user ID ever passed in.
- ▶ 35:46
Zero interest rates taught managers never to make technical decisions, leaving them only headcount and process as levers
Will was told as a new manager to “never make a technical decision again.” If you can't choose technology, he says, the levers left are headcount, scheduling and people, which drove bloated orgs and an overemphasis on hiring. Survey scores became goals rather than prompts for questions. Companies also let engineers skip on-call, though everyone knows owning on-call makes software better. Will's rule now: hiring is the least interesting way to solve any problem.
- ▶ 48:35
Staff engineers should work on what the business needs, not on what is fun
Will wrote Staff Engineer partly because he was frustrated with the idea that staff engineers get to pick interesting problems. At Stripe his strongest engineers asked for more interesting work, and his answer was “I only have suffering to offer” that will make the business succeed. The most senior people everywhere he worked did the valuable work the company needed. Will says the industry has now reverted to that view.
- ▶ 58:15
Staff-level work is navigating ambiguity and making decisions, and agents don't do that well
At Stripe, engineers leading a multi-region project wanted product to decide cross-region replication trade-offs under the CAP theorem, which product can't make. Will says staff and principal engineers should make trade-offs, build conviction and prototype to test assumptions. That takes knowing users and unit economics. He doesn't think agents are good at it, because the implicit context sits outside the context window. He also distrusts overly precise career rubrics, since people play to them.
Give your email to unlock the rest of the key points — for every episode.
Do more with Charlie
Free — sign in with your email, no password.