Field notes

Small signals from big systems.

Short observations from critical-environment operations, engineering, Lean and the AI layer. Less blog, more things noticed while the systems are running.

A coffee grinder imagined as a critical data centre machine

Field note · 8 September 2026

The terrifying sound of an empty coffee grinder

For years, my nervous system has been trained to react instantly to abnormal sounds: a bearing beginning to fail, a motor under heavy load, a pump starting to cavitate, a contactor chattering or arcing, or water running somewhere it definitely should not.

Spend enough time around data centres and machinery and you develop a sixth sense for abnormal. This morning, I heard a familiar motor suddenly accelerate under no load. My brain went straight into fault-finding mode: That RPM is wrong. Where has the load gone? Shaft snapped?

It was the office automatic coffee grinder. The bean hopper was empty. Fright, flight or freeze? Mostly coffee deprivation.

It did remind me, though, how much expertise lives below conscious thought. Experienced people do not always need an alarm or dashboard to know that something has changed. We hear it, smell it, feel it. We walk into a room and know something is not right.

That is pattern recognition built from thousands of ordinary moments: the biological AI model quietly running in the background. Apparently, mine is now being retrained for office-based critical infrastructure.

#DataCentres #Engineering #Operations #FacilitiesManagement #Coffee

Join the conversation on LinkedIn

Planned

Why RCAs go stale, and the control that fixes itDraft

Staleness is not a discipline problem, it is a design problem. What a problem management KPI framework has to measure if you want records that close.

Planned

The handover gap: how good builds become bad operationsDraft

The real issue is not only the build quality. It is whether the operational transition is ready: the documentation, the readiness checks, the training, the ownership model, and the customer confidence that the facility is truly ready to run.

Planned

Platform migrations: the stabilisation period nobody budgets forDraft

Lessons from taking a region through a ServiceNow transition. The cutover is the easy part; the ninety days after it decide whether the new model holds.

Planned

The most automatable work in a data center is not in the data hallDraft

Triage summaries, RCA drafting, knowledge maintenance, change risk assessment. Where AI earns its place in operations, and the three places it must never sit.

Planned

What a poultry processing plant taught me about hyperscaleDraft

Throughput is throughput. The constraint theory that works on a processing line works on a rack deployment pipeline, and it is more visible when you can see it move.

New pieces land here first. For anything in the meantime, the conversation usually starts on LinkedIn.

Open to conversations

Working on a similar problem?

Happy to compare notes on operations design, service management or AI-enabled workflow.

Get in touch See the track record