Insights, opinions, and deep dives into engineering excellence, technical leadership, and building software that lasts.
Most engineering managers learned to run 1-1s by defaulting to what they already knew: the work. But the sprint status is available somewhere else. The 1-1 is the only recurring instrument a manager has for actually knowing the person in the room.
Read the Full StoryA legal deadline is handed to you. A contractual one you agreed to — which makes it not a planning problem but a strategy problem wearing a planning problem's clothes. The fixed-scope, fixed-date contract feels like risk management. It's risk concentration, committed at the moment you know least.
Read the Full Story"Hard things first" is a motivational cliché. The disciplined version is sharper and stranger: confront the part you understand least, not the part that's most laborious — because uncertainty is only cheap to be wrong about while you still have time.
Read the Full StoryA self-imposed date on a big idea is a guess you've chosen to be held hostage by. A real, external deadline is something else entirely — a constraint you engineer within. The discipline doesn't break. It inverts.
Read the Full StorySetting a date on a big idea feels like a commitment. It's a guess wearing formal clothes. The discipline that actually produces calm work runs the other way around — fix the time, size the problem to fit, and let the team learn its way to the answer.
Read the Full StoryThe bottleneck was never the code. Agile, XP and DevOps said so out loud, decades ago. The companies discovering it now aren't learning something new — they're confessing they never understood the methods they claimed to run.
Read the Full StoryYou don't stay calm under pressure. You prepare to. Panic is what happens when the fundamentals you never made automatic start competing for attention with the problem that's actually new.
Read the Full StoryThe honest answer to most predictions about AI and engineering is "we don't know yet." That answer is harder to publish than a hot take — and more useful than one.
Read the Full StoryThe textbook Incident Commander doesn't exist on most teams — and pretending otherwise makes incidents worse, not better. How small-team commanders run the response with their hands in the work.
Read the Full StoryA small tool I built to stop leaving secrets in plaintext .env files. AWS credentials, API tokens, anything — encrypted at rest, injected into the subprocess on demand, gone when the process exits. Public, MIT-licensed, on GitHub today.
Read the Full StoryThe Lookout role pays for itself, but only if you build the rotation that holds it. How to give slack a name, a deliverable, and a person — without it becoming a luxury or a burden.
Read the Full StoryThe job of paying attention to the system — the role most teams have but never name. By the time something appears on a dashboard, it has been happening for a while. Someone needs to see the smoke early.
Read the Full StoryMost teams stop at resolution. The ones that don't are the ones that get better.
Read the Full StoryThe incident response foundations held — but everything around the edges fell apart. One unassigned role explains why communication chaos compounds during every incident.
Read the Full StoryMost teams handle small incidents fine through luck and proximity. But the same pattern that works for a five-minute fix produces hours of downtime when complexity doubles. Three structural changes cost nothing and change everything.
Read the Full StoryEveryone talks about speed. But the real shift isn't how fast engineers work — it's what they can now attempt. The execution tax disappeared, and the binding constraint moved upstream where it always belonged.
Read the Full StoryPanic is not a strategy. It is the absence of one. Urgency is a feeling. Clarity is a choice. And sustainable pace is not a compromise but a competitive advantage.
Read the Full StoryQuiet leadership produces steady, compounding results. But steady and compounding doesn't make the news. So the examples exist — they're just buried under the noise of everyone else performing urgency.
Read the Full StorySix words that replace every uncomfortable capacity conversation. Not "no," not a lecture on velocity — just a question that makes the trade-off visible and puts the decision where it belongs.
Read the Full StoryFive years at a startup that looked right on paper. The mission was real, the people were talented, and yet something was off. Ron Westrum's research on organisational culture finally gave it a name.
Read the Full StoryThe pair model solved the ownership problem. It didn't solve the resilience problem. For that, you need one more person — and what emerges is more interesting than just availability.
Read the Full StoryPairing changes who holds context, who makes decisions, and who feels responsible. But it also reveals something you can't see in a ticket-based model: who your leaders actually are.
Read the Full StoryThe dispatch model works brilliantly at 5 to 8 engineers. But past that point, the seams start to show. Here's how to recognise the inflection point and what comes next.
Read the Full Story