README

Most of the behavior you want to change in an organization is working exactly as designed. Nobody designed it on purpose. Somebody added every gate and every handoff locally, solving a real problem in front of them, and nobody was watching what those decisions added up to. That belief sits underneath most of how I lead, and it keeps redefining the job faster every year.

This is a document about how I work, written for the people who work with me and who I work for.

How I read an organization

When something is going badly, I start with the structure rather than the people. A team that adds checkpoints faster than you can remove them is usually responding to something rational; the last name on a change before it breaks in front of customers is the person who carries the exposure. If you simply remove the checkpoint, the exposure that built it stays where it was. Ask what people are protecting themselves from, and why that protection makes sense inside the system they’re standing in. Asking what’s slowing everyone down just gets you a new name for the same checkpoint.

A career ladder written as an HR formality ultimately decides who gets promoted and who leaves. Draw team boundaries in a reorg and they become the boundaries in the systems between those teams, long after everyone forgets the reorg. Nobody files either one as architecture, and both are. They’re the operating system, and they keep running whether or not anybody maintains them.

Every job description wants a systems thinker, and the label tells you almost nothing. Anyone can learn the vocabulary, and talking fluently about second-order effects costs nothing. I watch the size of the boundary a person draws around their own problem — whether it matches the real system or just the shape of their role. You see it in what someone did the last time their area broke, and in whether they noticed something outside it that nobody asked them to.

Whether someone can put an idea in the shape a particular audience will carry matters to me about as much, and that has nothing to do with being the loudest person in the room and everything to do with whether the idea survives leaving it. So does whether they leave the people around them more capable than they found them. Neither shows up cleanly on a résumé, and both change what a team can do more than individual output does.

What I think the job is now

For most of our careers until now, understanding a system narrowly was safe enough. Two things capped how far the consequences could travel — the edge of your job, and the plain expense of execution. Even someone who understood everything couldn’t act on more than a slice of it. Both caps are gone. One person can act across an entire system directly, and when understanding doesn’t stretch as far as reach does, change outruns anybody’s grasp of how the pieces fit.

That moved the hard part of the job. Ownership was always layered — completing the task, standing behind the outcome, and holding the conditions that produced the task — and those layers are separating faster than most managers have adjusted to. Finishing the task is the most visible layer and the least interesting, and it now costs almost nothing. I watch for ownership that presents well and holds nothing underneath, where someone signs off on work they can describe but can’t explain the reasoning behind. Real accountability means saying out loud what you verified yourself and what you took on faith.

I also don’t run technical investment as a separate budget. Reserve a fixed percentage of a team’s capacity for infrastructure and you set up a negotiation the technical side loses slowly. You also teach everyone that the platform is a cost center bargaining against the business. Every technical investment is a business decision carrying a return and a cost of delay, and I’d rather argue it on those terms than defend a quota nobody outside technology believes in.

How decisions get made around me

My input is a hypothesis unless I say otherwise. When I react to your plan, you’re learning how it lands on someone who wasn’t in the room while you built it. Weigh it at roughly that value. You keep the decision, and telling me where I’ve got it wrong is part of making it.

When I state a strong preference, I give the reasoning and then explicitly release it, because a preference I repeat often enough and never release becomes a rule nobody remembers agreeing to. When a question belongs to whoever owns that domain, I hand it over and take their answer above my own instinct. When you answer a challenge of mine well, the challenge is done; a good answer was the point of raising it. When a disagreement won’t resolve, I put my concern on the record once and get out of the way. And when I’m arguing for something I’m not the one who pays for, I try to say so before I make the case.

Sometimes the call is genuinely mine. When that happens I make it in the room, in one sentence, with the reason attached, and I try to keep it rare enough that you notice it happening. A leader who wins on authority stops being told things, and a leader nobody has ever watched lose an argument stops being argued with.

What I hold a bar on

A standard I can’t describe isn’t a standard, just a preference I get to apply unevenly. I’ve argued in public that taste, the way most people invoke the word, is inarticulation with better branding — you can’t hire for a bar you can’t define, you can’t develop anyone toward it, and it stops where the person holding it stops. So I owe you the written version of what I actually expect.

Anything that makes a decision records why it decided that, with what inputs, under which version. Nobody can debug or defend a system that can’t reconstruct its own reasoning, and that gets expensive the first time someone challenges an outcome. Design the failure path before the happy path. List what goes wrong before you write any retry logic, and give every external call a timeout and a fallback. Fail loudly rather than quietly. We should never find out something is broken from the person it broke for.

Name the owner before you build the thing. Anything you stand up is something somebody now has to secure and monitor for as long as it exists, and the people who inherit that work are rarely the ones who proposed it. The bar isn’t whether it’s useful but whether it’s useful enough to own permanently. And having access to data doesn’t mean we may use it. Consent and stated purpose decide what moves and who we’re allowed to contact. I’d rather build that boundary into the schema than write it in a policy document, because people forget policies and schemas don’t budge.

How information reaches me

I don’t chase updates. Keeping me informed is part of the work, and we’ll agree explicitly on what that means — what belongs in a written weekly, and what needs me the same day. If I’m asking you for a status, the system we built has a gap in it, and I’d rather fix the system than repeat the question.

The signal I watch hardest is quiet. Credibility doesn’t erode on the day people push back on me. It erodes on the day they stop bothering, and by then it looks like a standard everyone politely routes around. If something I hold isn’t worth your time, I should either make it worth your time or retire it. Tightening it only proves the point that made people start avoiding it.

Working with me

Our 1:1s are yours. I’ll bring things when I have them, and the default agenda is whatever is actually on your mind, including the thing outside work that’s making this week hard.

You won’t have to guess where you stand. I give feedback close to the thing it’s about and in specifics, and I say the appreciative kind out loud instead of saving it for a review cycle.

I’ll share as much context as I have, including the reasoning behind decisions you didn’t get a vote on. When I can’t share something, you’ll get a plain “I can’t share that” rather than a carefully constructed non-answer. And I’m interested in where you’re going, not only what you’re delivering now, including the version of that where you outgrow working for me.

The standing invitation

I revise this once a year, most recently in August 2026. If something here doesn’t match how you’ve actually experienced working with me, that gap is the most useful thing you could bring me.