18F — Eight Years Inside the Machine

What I walked into

I arrived at 18F in 2017 thinking I understood how government worked. I didn’t. I understood how projects worked. Government isn’t projects. It’s procurement cycles that outlast any roadmap, hiring freezes that arrive without warning, political transitions that rewrite priorities overnight. The org chart reflects a political reality, not a design logic.

18F was supposed to be the startup inside government. In some ways it was. But the interesting question was never “how do we work around the constraints?” It was “what does good look like inside them?”

The people

Here’s the thing people don’t realize about jobs within TTS, especially at 18F: people across industries who work in tech knew about us. As civic tech’s lore grew, the place you go in government when you want to do good stuff or help people, it meant people actively sought out our job postings. Our postings would go up for a week, sometimes only a single day. Even with those restrictions, we’d consistently get 1,000+ applicants for every opening.

I was Head of Design for close to four years, overseeing what was the largest in-house team of human-centered designers anywhere in the federal government. When we finally got the green light to backfill attrition, it let us focus on priorities I cared about: growing our UX engineering and product design expertise, expanding our content strategy outfit (still the best in the entire government) and most importantly, building the first in-house service design unit anywhere in government.

Our staff represented the country. I had colleagues from Alaska to rural western states like Montana, from Puerto Rico to the Deep South. Far more people living on farms in rural states than in big cities. This felt right for work that impacted people everywhere.

Public service isn’t a consolation prize. Every single designer we hired was top-tier. Even our backups, and the backups’ backups, would’ve been fantastic hires. It might surprise some people that public servants genuinely care about their neighbors, even those they’ll never meet. People want to contribute to something bigger than themselves.

The visibility problem

The deeper I got into leading teams, at 18F, in teaching, in coaching, the more I kept running into the same breakdown. Delivery is a visibility problem. The frameworks we use to ship things assume shared context will naturally emerge. It doesn’t. Someone sees a performance issue, someone else sees a scope problem, leadership sees a timeline risk. Everyone’s solving a different problem, and nobody realizes it until the retrospective.

In my tennis coaching, early on I noticed players looking at me in crucial moments when their eyes should’ve been on the court. I vowed to instill a level of confidence in them that even when their decisions don’t work out, they can learn to trust themselves. On changeovers, I’m very often saying “I trust you.” That instinct, trusting the person closest to the play to make the call, is the same thing that makes delivery teams work.

The best teams I saw at 18F could do three things: describe what they were doing and why, question assumptions without fear, and recover quickly when something changed. The struggling teams had plenty of talent. What they lacked was the ability to see the same picture at the same time.

Because 18F was designed so people would rotate out on term-limited appointments, there were defined rituals around documentation and knowledge transfer. We had rituals for surfacing what we were seeing, beyond just what we were doing. Authority rotated based on who had the clearest read, regardless of seniority. We rebuilt the picture constantly because teams were always changing and we couldn’t afford to let context drift. That level of clarity made it easier for new people to join without losing rhythm.

The political part

In senior leadership at TTS, the parent org of 18F, login.gov, and cloud.gov, I got a different education. Translating between technical teams and political leadership. Managing communications during uncertainty. Coordinating strategy across products with different funding models, different constituencies, different definitions of success.

The hard lesson

You can build something excellent inside an institution and still watch it get dismantled for reasons that have nothing to do with quality. 18F closed. The work didn’t disappear. It lives in the agencies it changed, the people it trained, the practices that became standard. But the institution couldn’t survive the politics.

I’m not worried about the people we hired finding new opportunities. I’m disappointed they won’t be able to work on the things they cared about. I’m disappointed for the people who won’t get a chance to learn from them. And ultimately, I’m disappointed for the country.

Institutions remember practices, not people. The brilliant solo contribution gets forgotten. The hiring process, the critique protocol, the way you documented things. Those survive. If you want lasting impact, build things other people can maintain when you leave.

Eight years taught me that making institutions work differently is slow, unglamorous, and deeply political. It’s also, and I didn’t expect this when I started, the most important work I’ve done.