Skip to content

The People Your Process Can't See

Published: at 11:54 AM
0 views

I used to treat pre-dawn merges as a productivity story—someone got into flow. Then the pattern held: fewer incidents, quieter on-call, smoother releases. Somewhere in the system, a person was quietly lowering entropy.

I finally asked. He pointed to life happening at once: a breakup, a parent in the hospital. Work had become the place where he could move grief into motion. He never asked for anything.

He left soon after for a team that noticed without being prompted.

Every org has these quiet load-bearers. They operate in the interstices of your charts—across interfaces, across time zones, across job descriptions. They are the missing bolts you notice only when the structure shudders.

They don’t map to ladders that assume one lane. One week they’re de-flaking CI to give everyone an hour back. The next, they’re rehabbing a migration plan so Support can breathe. They coach a new hire through on-call, then write a customer note that keeps a renewal alive.

Scorecards struggle with this kind of contribution.

When your best operators start keeping receipts—threads, decks, self-reviews—take it as a warning. They’re not naturally self-promotional; they’re adapting to a system that confuses amplification with value.

At almost every high-growth company I’ve joined, the person no one could replace was the person no one could promote. Not for lack of ability, but because their work cut across boundaries the calibration grid didn’t know how to credit.

What this looks like up close:

When they move on, you find archaeology: a midnight runbook; a small service shielding you from a vendor outage; a spreadsheet that somehow became a team’s nervous system; quiet relationships that bridged gaps titles never acknowledged.

Everyone who stays learns the lesson: essential and recognized are not synonyms.

If you want to keep them, change the system, not the person.

  1. Make drag reduction measurable.

    • Count toil removed, risks retired, and coordination created. Put it next to feature count.
  2. Create a track for systemic impact.

    • Build roles that reward making the whole system better—reliability leadership, design systems, platform stewardship.
  3. Compensate without the ask.

    • Equity and raises should meet cross-boundary impact proactively. If they have to lobby, the trust has already slipped.
  4. Institutionalize credit.

    • Keep a “glue work” ledger, a reliability changelog, a place where prevention is documented and celebrated.
  5. Tend the human.

    • 1:1s that ask about load outside work, not just roadmaps. Offer relief before burnout becomes the coping strategy.
  6. Normalize boring wins.

    • Celebrate quiet stability the way you celebrate launches. Make calm a metric.

Recognition is timing and attention, not a spotlight. Sometimes it’s a title that doesn’t fit a template. Sometimes it’s a raise that lands without a negotiation. Sometimes it’s a single line in a meeting: “This work changed our trajectory.”

Your job isn’t to fit them to a level. It’s to see them in time—while they still believe being seen here is possible.

When the activity graph shifts to odd hours, resist the reflex to open the diff first. Open a conversation. Ask what would make next week lighter. Ask what future they want here.

If you only count what’s loud, you’ll miss what’s holding you up.


Previous Post
Inverse Scaling Laws in Neural Networks
Next Post
Why Complex Systems Can't Be Designed