In partnership with
Good tools answer the question for you, until the one they can't. Curiosity is a muscle, and this is how to keep it warm on purpose.
In this drop
The point: Every routine question your tools answer is a question you stop asking yourself, and curiosity is a muscle that fades when it is not used.
Why it matters: The cost is invisible until a genuinely novel failure arrives and finds you out of practice.
Try this next week: Spend twenty minutes on a system that just works, asking 'and then what?' of its telemetry and pulling one thread to the end.
The point
Good tools answer the question so fast you forget you were the one who asked it. The dashboard shows the number, the runbook gives the step, and now AI has read the incident and summarised it before I finish the first line. Every one of those is genuinely useful. I am not here to complain about the tools.
But I want to be honest about something I have noticed in myself. I have stopped looking at the systems that just work. No incident is driving me to look, so I do not. The tool answers the question and I move on.
Curiosity is a muscle. The moment your systems stop requiring it, you stop exercising it. Every routine question the tool answers is a question you no longer ask yourself. Convenient, and slowly corrosive, because you are not keeping up the habit of asking the next question.
Reality check
The number to watch is not how well you handle the incidents your tools explain for you. It is how well you handle the one they cannot.
One proof
Human-factors researchers have a name for this. Raja Parasuraman and Dietrich Manzey, in 'Complacency and Bias in Human Use of Automation' (Human Factors, 2010), describe how reliable automation pulls attention away from the monitored process, so the operator notices less and investigates less, precisely because the machine is usually right. The skill that decays first is not the hands, it is the questioning. Field note, general and not a benchmark: the engineers I have watched sail through a genuinely novel incident are always the ones who kept interrogating the quiet systems when nothing forced them to.
Where this breaks
This breaks if you read it as an argument against AI or automation. It is not. Automation, mature tooling and plain seniority are all worth having, and all quietly retire your questions. The failure is letting them retire the habit as well as the task. Keep the tools; put the reps back on purpose.
Try this next week
Pick one system you never look at because it just works. The stable platform, the boring bit of hardware.
With no incident forcing you, take twenty minutes, open its telemetry, and ask 'and then what?' of the first thing you find.
Chase that thread all the way to the end. That is one rep. Do it before the night you need it, because you do not want to be relearning this at 3am.
Signal Check: your questions answered
A new part of the show and the newsletter. Your real questions, straight answers. Two from this week.
How do you find the change behind an incident without reconstructing it from every log and alert timestamp? Start from the change, not the symptom. Line the deploy up against the moment the behaviour actually shifted, and let the telemetry tell you which one fits. Do not stop at the first correlation. A deploy that landed is a suspect, not the verdict. The teams that make this painless have wired their change markers into the same view as their signals, so the question half answers itself, and they still interrogate it by hand.
Is AI actually improving IT operations, or just adding another layer of dashboards and noise? It is one driver among several, not a revolution and not a fraud. It helps when it shortens the distance to the first question, a summary, a likely cause, the first thread to pull, and when it surfaces a pattern you would have taken ages to find. It hurts when you let it be both the last and the first question. Let it retire your first question if you like, but never let it retire the first one.
Three links I am watching
Lisanne Bainbridge, 'Ironies of Automation' (1983): the classic on why automating the routine leaves the human least practised at the moment it matters.
Google SRE, the 'Wheel of Misfortune' incident role-play: a light way to give a team reps on failures that have not happened yet.
Signal Drop 22, 'The Six-Week Decay': the companion on how automation de-skills incident response.
Which system have you stopped looking at because it just works, and what would you find if you spent twenty minutes asking 'and then what?' of it tonight? Reply and tell me. The best ones become a Signal Check.
Allan
|
Two ways to go deeper The book Metrics & Mayhem: A CTO's Guide to Observability That Actually Works. Kindle, paperback and hardback are out. Free chapter · Get the book Work with me A fixed-scope Observability Assessment for senior IT and engineering leaders. See how it works |
4x your communication output. Same quality. No burnout.
The bottleneck isn't what you want to say — it's how long it takes to type it. Wispr Flow removes the bottleneck.
Speak naturally and get polished, send-ready text for executive summaries, client updates, board recaps, investor notes, or just the 30 Slack messages you're behind on. Flow strips filler, formats numbers and lists, and preserves your tone.
Used by teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works in every app on Mac, Windows, and iPhone.
PS: the full episode, including this week's Signal Check, is on Spotify and Apple Podcasts.