In 2023, my organization stopped functioning. Not gradually, but with the suddenness of a system hit by a cascade of unbuffered change.
We had just absorbed several acquisitions, each bringing its own definition of urgency. Our engineers were drowning. TOIL—the repetitive, manual, interrupt-driven work that erodes engineering value—climbed to a staggering 83.9%. We were running constantly, yet nothing was moving.
This collapse was particularly painful because it followed years of hard-won progress. Each prior merger had been absorbed faster than the one before—two years, then one, then six months. The framework was working. Then it wasn’t. We didn’t get there by shipping a new observability stack or adopting a trendy incident framework.
We did it by rebuilding the thing that sits between our engineers and the chaos of the outside world. It is a concept most SRE teams never explicitly name.
I call it the Membrane.
The Fiction of the Org Chart
Most organizations view hierarchy as a safety net. They are wrong. Niklas Luhmann, the sociologist and systems theorist, correctly identified that organizations are not pyramids of power; they are systems of communication defined by their boundaries.
In the high-stakes world of SRE, the org chart is fiction. Hierarchy tells you who reports to whom, but the membrane tells you what the organization actually lets in—and therefore, what the organization actually is. To survive, you must stop building silos and start building membranes.
A silo is a wall; it’s impermeable, creates bottlenecks, and fosters “not my problem” cultures. A membrane, however, is a semi-permeable filter. It separates essential signals from debilitating noise. Gatekeeping isn’t a bureaucratic hurdle designed to slow people down; it is a life-support system. It shields builders from distraction while remaining permeable to genuine, validated needs.
A membrane is not a single gate. Systems maintain identity through boundaries—plural, each with its own calibration. Some filter noise; others rotate people, govern partner accountability, or absorb mergers. What follows describes the first.
Your Intake Board as an X-Ray
At our core, we implement this through visible intake boards where triage criteria function as the mechanical settings for permeability.
Your intake board is not a productivity tool. It is an x-ray of your membrane. A team whose intake board looks like a parking lot of stalled cards has a membrane that is too tight. A team whose intake board looks like a firehose has no membrane at all. Neither team is failing because of their ticketing tool. They are failing because no one has taken responsibility for the mechanical settings of the filter—the triage criteria that decide what gets through, in what form, and to which person.
This is where we embrace the “Olivetti” perspective: team performance cannot be measured by a throughput index alone. Adriano Olivetti understood that a team is a community to be cultivated, not a resource to be optimized. Burnout prevention is a moral imperative, and the membrane is the architecture that makes that cultivation possible. By defending an engineer’s attention, we are defending their dignity and their ability to do deep, meaningful work.
The 2023 Breach: A Lesson in Calibration
The membrane is a living thing that requires constant tuning. Our 2023 crisis occurred out of unforeseen circumstances.
As we integrated new acquisitions, we attempted to absorb new products and cultures—with their undocumented tribal knowledge and manual processes—without re-calibrating our filters. The result was a breach of our operational integrity. We had to step backward in maturity. The frustration was palpable: We had solved this before; why were we solving it again?
The recovery took us through 2024 and into 2025. The membrane framework didn’t prevent the problem, but it allowed us to metabolize it. We used the 83.9% TOIL peak as the data input required to re-tune our filters. Under Google’s strict 5-point TOIL definition, we drove TOIL from 59.7% in 2024 to 44.7% in 2025 — back below the SRE health benchmark. We compressed our P95 cycle time — the true pulse of an agile organization — from a glacial 294 days in 2020 to just 57 days in 2025. It proved a vital principle: an uncalibrated membrane is effectively non-existent.
The Engineering of the Boundary
The SRE industry has spent a decade perfecting the “inside” of the membrane. We have excellent observability, automated runbooks and blameless postmortems. The craft at that layer is mature.
But the boundary itself—what comes through, what gets sent back, who decides—is often treated as “soft” work. We dismiss it as “people stuff” or office politics. I have found that dismissal to be incredibly expensive. Treating the boundary (or filter) as anything less than a first-class engineering problem is how teams drown.
I challenge you: Open your intake board tomorrow morning. Look at it not as a list of tickets, but as a live x-ray of your membrane. Ask yourself:
- Which request did you let through this week that failed the triage criteria?
- What did we block that should have been an urgent escalation?
- Who paid the price for that calibration error, the engineer, or the requester?
- Are we defending systems or enabling teams?
If the answer is “I don’t know,” you have found your next engineering project. Calibration is not “extra” work; it is the only work that ensures your system survives.
