Stay Out of the Way: The VP Engineering Job Description Nobody Posts

Every VP Engineering job description reads like this:

“The VP of Engineering will drive the technical vision, own the engineering roadmap, partner cross-functionally with Product and Design, deliver a high-performing engineering organization, lead through strategic initiatives, build and scale a world-class team, foster a culture of innovation and operational excellence, and align engineering investment with business outcomes.”

That paragraph is a verbatim synthesis of ten real public job listings. Every verb in it appears multiple times across the corpus. Scale AI, Dow Jones, DriveWealth, Klue, DEFCON AI. The recruiter templates. The same eight verbs, assembled in slightly different order. Drive. Build. Lead. Scale. Foster. Align. Deliver. Own.

That paragraph does not describe my job.

Here is a Tuesday from my actual calendar:

An architecture review where the IC’s proposal was sharper than mine. I said so, and walked out. A standing meeting series I cancelled because nothing decided in it could not have been decided by the senior engineers on the team without me in the room. A Sales request for a customer commit that I declined and pushed back on with the actual delivery math.

None of those are on the JD. They are most of the job.

The JD describes a different role entirely. Here’s what the corpus actually shows.

The gap is not marginal

Ten real public VP Engineering listings, spanning early-stage to public company. The verb count: lead (18), build (14), drive (12), scale (9), mentor (9), define (8), align (6), deliver (6), develop (5), own (5), foster (5), partner (5). 129 additive verbs total.

Against that: verbs that explicitly name subtraction (decline, kill, cancel, remove, push back, defend, protect) appear zero times across the corpus. Zero. Verbs that carry semantically subtractive meaning (prioritize, focus, balance, tradeoffs) appear five times across the corpus. The ratio at the most charitable read is 129:5. The additive framing does not compete with the subtractive framing; it crowds it out almost entirely.

Two forces explain the gap. First: HR and recruiting know how to put additive work on a JD. Deliverables, outputs, artifacts you can point to. Subtraction is invisible. Nobody lists “cancelled a meeting that shouldn’t have existed” in a job posting. The JD apparatus records the shape it can see. Second: the informal startup doctrine. Martin Casado’s 2017 a16z post on hiring a VP Engineering has shaped how an entire generation of founders thinks about the role: active builder, culture-setter, chaos-handler. Casado’s framing captures one half of the job: the parts that are visible to a JD. It misses the other half.

Andy Grove’s foundational frame makes the problem visible: “A manager’s output = the output of his organization + the output of the neighboring organizations under his influence.” (High Output Management, Vintage Books, 1995.) If the VP’s output is the organization’s output, then any VP action that does not improve organizational output is waste. The JD’s additive verbs describe activity; Grove’s frame measures results.

Emily Nakashima, in “On Becoming a VP of Engineering, Part 2” (Honeycomb, August 2024), is the closest explicit articulation of the subtractive thesis in the practitioner literature. She names the structural invisibility directly: “The further you go on the management ladder, the more doing a good job in your role can become decoupled from appearing to your team to do a good job in your role.” That decoupling is what a JD cannot capture. Camille Fournier (The Manager’s Path) and Will Larson (An Elegant Puzzle, The Engineering Executive’s Primer) are consistent with the thesis but neither states it explicitly.

There is precedent for the argument at the IC level. Tanya Reilly’s “Being Glue” (noidea.dog, 2019) names what happens when invisible work goes unrecognized: “We lose good engineers because they happen to also be good at other skills we need.” Invisible work that determines team success appears in no JD. This post extends that argument two levels up.

Get out of the way of better ideas

Most of the job is subtraction. The hardest part is getting out of the way of better ideas when an IC has one.

I had directed execution on an architectural decision recently. The instinct was clean, defensible, the kind of call a VP is supposed to make. An IC offered insightful pushback: a sharper proposal, better-fitted to the constraints, supported by data the IC had spent more time with than I had. I immediately recognized the better proposal, accepted it, and stepped away to let them handle the rollout. The team executed cleanly. The moment is not remarkable because the IC was right. It is remarkable because I didn’t override.

What the JD would say if it described the actual job

Here is what the JD would say if it described the actual job.

The VP of Engineering will remove blockers from a team of brilliant people, decline most of the work that lands on the team’s plate, intervene only when strategic drift, hire/fire, executive misalignment, cross-team friction, incident escalation, or external commitment crosses the bar that requires this seat, and otherwise stay out of the way.

The VP will cancel meetings that produce no decisions the team cannot make without them. Kill projects that have passed their useful arc. Hold the line on hiring quality. Push back on Sales commitments that misread delivery capacity. Get out of the way of a better idea when an IC has one.

And when a situation crosses the threshold, the VP will intervene fully: make the structural move, not the gentle nudge; name the performance problem directly and authoritatively; absorb the political friction that only this seat can absorb.

Then back away again.

The honest JD is longer because it has to itemize the work the composite condensed into single verbs. That is the point. The composite verb “lead” is one word; the honest decomposition (cancel meetings that produce no decisions, decline customer commits that misread delivery capacity, get out of the way of better ideas when an IC has one) takes a sentence each. It contains all the subtractive verbs the composite JD was missing: remove, decline, cancel, kill, push back, intervene. That is not a stylistic preference. It is a description of the actual job.

What the team is paying for

A CEO funding subtractive work has a verification surface, even when the calendar looks like nothing. Five metrics show whether subtraction is producing throughput, quality, predictability, reliability, and decision-distribution: cycle time, rework rate, velocity predictability (the variance is the signal, not the average), incident MTTR, and escalation rate (the ratio of decisions that escalate to the VP seat versus resolve in-team). These are observable, falsifiable, and longitudinal. The team is paying for movement on these signals, not activity on the VP’s calendar.

The visible job and the actual job

129 verbs of additive work across ten real public VP Engineering listings. Five verbs of even charitably-subtractive work, scattered. That is the visible job, written by people who can only describe what they can see.

The actual job is the inverse. Most weeks are mostly subtraction: removing blockers, declining work that does not belong on the team’s plate, cancelling meetings, getting out of the way of better ideas when ICs have them. The Tuesday calendar is more typical than the JD.

The most visibly active VPs are often the most expensive. The teams that fund them are paying for activity on the calendar, not movement on the metrics.

References

  • Casado, Martin. “Hire a VP of Engineering.” Andreessen Horowitz, May 26, 2017. The standard additive doctrine: VP as active builder, culture-setter, chaos-handler. The post engages this framing as the informal startup doctrine that filters into JDs via recruiters.
  • Grove, Andrew S. High Output Management. Vintage Books, 1995 (orig. Random House, 1983). The foundational frame for managerial leverage. The cited passage on managerial output grounds the claim that VP activity unconnected from organizational output is waste.
  • Nakashima, Emily. “On Becoming a VP of Engineering, Part 2.” Honeycomb Blog, August 2024. The closest explicit articulation of the subtractive thesis in the practitioner literature. The cited passage on the structural decoupling between doing a job well and appearing to do it well is the post’s load-bearing canon citation.
  • Reilly, Tanya. “Being Glue.” noidea.dog, 2019. The IC-level precedent for the structural-invisibility argument this post extends to the VP seat.
  • Fournier, Camille. The Manager’s Path. O’Reilly, 2017. Scaffolding canon. Treats the VP level primarily through execution ownership rather than constraint-clearing.
  • Larson, Will. An Elegant Puzzle. Stripe Press, 2019. The Engineering Executive’s Primer. O’Reilly, 2024. Scaffolding canon. Both consistent with the subtractive thesis; neither states it explicitly.