Home

How I lead

Most of what I've learned about leading comes down to restraint — what not to do, what not to ship, and when not to step in.

I founded a game studio that Electronic Arts acquired and the payments company Spreedly, then led product and design at Observable, Tray, and CloudBees. Twenty years in, the lessons that held up are simpler than I expected when I started.

Direction before speed

A team moving fast in the wrong direction just reaches the wrong place sooner. Velocity multiplies against your heading, so a wrong heading makes every sprint more expensive.

I set direction first, clear the obstacles slowing the team, and only then push on pace. Most startup advice runs the other way. Speed is the easy thing to celebrate and the last thing worth optimizing.

Outcomes over output

A team can ship constantly and change nothing. Output is easy to see — tickets closed, hours logged, features released — so people drift toward the visible proxy and away from the result. Usually not on purpose. They optimize for the appearance of work because that's what gets rewarded.

AI raises the stakes. It can fill every proxy metric we have — ten times the output at no more real value, sometimes less once it all has to be reviewed and maintained. So I hold people to outcomes: did the customer's problem go away, did the business move. That means killing work that produced a lot and changed little, and trusting that the people doing the most valuable work often look the least busy.

Set the guardrails

Much of leadership is setting guardrails and tuning them. Too tight and you crush ownership and fresh ideas. Too loose and people flail. Where the line sits moves with each person and each problem — something you tune by hand, not a formula. Get it right and the team's results outrun anything you'd have produced alone.

Guardrails are also what make delegation real. Without explicit boundaries, every decision escalates back to the leader, because no one is sure what they're allowed to decide. So I'm clear about the mode I'm in — collaborating means we own the result together; delegating means I've handed ownership over and stepped back. The damage comes from calling something delegation while staying in every decision.

Hire for judgment, and know your gaps

When execution gets cheap, judgment gets scarce. AI can write the code, draft the copy, generate a hundred options — so the valuable person is the one who can tell which option is right. I hire for taste and discernment over raw speed, even though judgment is harder to interview for and harder to teach.

The same honesty applies to me. My job is to make sure the team is complete, which only works if I can name where I'm weak and hire judgment that covers it. The dangerous gaps are the unacknowledged ones, where my blind spot quietly becomes the team's. Mine is in the next section.

Feed the weaker engine

Every company runs on two engines — the one that builds the product and the one that takes it to market. It goes far only when both run strong. My own gap lives here: I pour energy into the build engine because that's where I'm most comfortable, then wonder why a strong product isn't growing. Naming it is what keeps me funding the engine I'd rather ignore.

At scale, change is a campaign

A big team carries momentum, and when you first turn the wheel nothing happens. That early resistance usually isn't a sign you're wrong — it's physics. The organization has been moving one way for years, and that direction is real investment and real identity for the people inside it. The two ways to misread the resistance are to give up, which teaches everyone that change is just noise to wait out, or to force it through, which buys you friction and burnout. What works is sustained pressure in a clear direction: keep turning the wheel, keep saying why, and accept the lag. Once the ship comes around, the same momentum that fought you starts working for you — the new way becomes the habit, then the culture.

I still build

I never stopped making things myself. At Field Bureau I design and ship a suite of local-first tools — a recorder that transcribes on device, a Markdown task manager, a library agent — built by a small pod working alongside AI. No servers, no accounts, no subscription; the files stay yours and the work outlives the tool.

I build them because I believe in that kind of software, and because it keeps my hands on the new economics firsthand. A handful of people can now hold an entire stack — native macOS and iOS, on-device ML, sync, design — and move at a pace that used to take a department. Leading well through that shift is hard to fake from the sidelines.

Judgment is the job

As machines absorb more of the execution, what's left for people is deciding what's worth doing and knowing when it's good. That thread runs through all of it — direction, outcomes, taste, the honesty to name your gaps.

Twenty years in, I measure a leader by what their team makes and whether it was worth making. Making sure the right things get made, by people better than I'd be alone — that's the job.

I write more about this at notsocommonthoughts.com.