lessons

26 posts, newest first

2026

10-28
Batch size, two years after DORA measured the AI paradox1min
10-16
Restructuring will not fix the queue1min
10-15
Developer experience research pays off in quality before speed2min
10-11
Ask the queue, not the team1min
09-15
Code review is the queue nobody staffs1min
09-12
AI's return on investment is a property of the delivery system1min
09-03
What actually drives developer experience1min
08-08
Trunk-based development: the evidence and the tantrums1min
07-31
Chat channels are queues wearing costumes1min
07-18
Little's law sets your lead time, not your roadmap1min
07-15
The meeting audit2min
06-27
On-call is a budget, and yours is being overspent1min
06-13
Change failure rate is the one DORA metric you cannot game2min
06-03
That famous 55 percent AI speed-up has an asterisk1min
05-18
A delivery overhaul, read as a worked example2min
05-04
Platform engineering pays rent in queue time2min
04-22
Goodhart's law has a bibliography1min
03-31
The productivity metric that became a target in public1min
03-21
The retro that repeats itself1min
03-06
Deploy frequency is a symptom, not a lever1min
02-21
The real cost of an interruption is twenty-three minutes1min
02-10
Productivity was never one number1min
01-21
The thirteen hours nobody budgeted1min
01-15
AI tools reward the teams that already did the prep work1min

2025

12-24
AI made coding faster and delivery slower1min
12-15
Diagnose the delivery system before you try to transform it2min