How To Leave

The opposite of running isn't staying. It's leaving well.

“I’m leaving.”

Nick was relating his latest psychedelic experience to the rest of the StartPlaying crew. “And so I’m sitting there, zoinked out, listening to Achilles Last Stand by Zeppelin, drool coming out of the side of my mouth. And that’s when I see the great grasshopper god…”

Devon looked up. “Whoa you haven’t even finished your drink.” He pointed at my mostly full pina colada, “Plus you were going to tell that story”

“What story?”

“The interview story. The bathroom story.”

“You mean the time that I aced an interview from inside the Google bus bathroom?”

Nick gave a long smile and patted on the chair beside him. “Come on, buddy. The night is young.”

I sighed. I was always terrible at fighting peer pressure. “Where do I begin?”

When I started at Google, I was ebullient. Growing up, it was the company du jour. Jam-packed with PhDs and prodigies capable of rewriting Dijkstra’s algorithm in their sleep. And now I was a SWE on Android with a massive data set spanning a billion plus devices. My fingers twitched with excitement.

In the first two weeks, I was even more overjoyed. My team hosted a hackathon, and I had just the idea: using the reams of columnar data to improve battery life predictions across all devices.

I pulled together a team of engineers whose eyes had lost the glint of soul. And we tirelessly worked to train and stand up a new model.

We won the hackathon hands down. Getting some silly prize of recognition and a set of steak knives. Afterwards, I met the VP to receive our prize. I was nearly bouncing off my chair.

“So when are we gonna deploy it?”

“Deploy what?”

“The model.”

He gave me a look that a father might give his son after the family dog has died, and began explaining OKRs. I felt my stomach drop.

I instantly looked for an escape hatch. I browsed other teams, and found one of particular interest: Semantic Annotation For Text. A team using old school LLMs, what were called LSTMs for those too young to remember, to research silly things like named entity recognition or Parsy McParseface. The team was eventually gobbled up by Google Brain, in turn gobbled up by DeepMind.

I formed deep learning paper groups, reimplemented backprop, started a data science YouTube channel and got my first million views. I read multiple papers a day, stacking them like magpie, making a tower so tall it prevented me from lowering my standing desk. I became avant-garde in that first little AI summer.

“I made this new experiment, doing adversarial search on LSTMs for word importance.” I turned my laptop around to show my new manager.

“Are we done with building the previous pipeline?”

“Yeah, finished that last week. I was wondering if we could run this experiment…”

She gave me a look I knew too well and began explaining OKRs.

I stared back, stunned into silence.

The problem never was my team. And if instead of fleeing, I talked to my friends and coworkers about it, I would’ve known the problem was big companies. But I didn’t. I just ran.

“By the end of it, I hated every day there. Slinking to work. Composing SQL queries so boring they’d make Claude want to segfault itself.”

The mood suddenly felt dark. We were supposed to be laughing. Telling a final joke before the night was over. Avoiding the question that was on everyone’s mind.

“Is that why you left Startplaying?” Nick said, staring down into his drink.

“No. I’ve never been happier than being with y’all and spending my time at Startplaying and experimenting and talking to users and watching crowds of geeks go up to Devon for pictures. Having one of our users get a tattoo of us. Supporting thousands of the best job ever and building this thing from the ground up and proving that one friend wrong that dissed our company without knowing I was the one behind it and growing with all of you and fixing mistakes.

“Nick, I didn’t leave because I was unhappy here.”

“Then why…?”

Y is the worst letter in the greek alphabet. It’s an accusation and a regret wrapped in a monosyllable. Damn the English language condensing its wherefore.

“Why” is an overspecified question. Why did the Roman Empire fall? Why do we tell stories? Why did Greedo shoot first? There’s never one answer.

Why did I leave? I left because the AI revolution will indelibly shape the world. I studied data science and machine learning before it was cool, and now all of a sudden it’s become the hottest thing ever. I left because I built what I wanted to. I left because I honestly believe that Devon will do a better job at leading Start Playing than I will. I left because, I, like everyone forever and for all time, believe that I won’t make the same mistakes again. I left because I needed to know if I could. I left because it’s what my heart wanted. And you can’t ignore your heart forever. I left because it was the right time to leave.

But that’s not what I said.

When I left Google, I picked up this book, Build by Tony Fadell, one of the original instigators of the iPhone and the creator of Nest. Build is a bildungsroman. It’s a coming-of-age story of a young techie trying to figure it out. He doesn’t run through his successes like a bloviating braggart composing a twenty-tweet thread. Instead he dwells on his foibles, reminisces, indulges a touch of nostalgia.

He talks about when he left his first company. He didn’t flee. He gave his notice. He tidied things up. He left things better than he found it.

I have never done that before, not really. I’d shyly give my notice and run away, trying not to overthink or dwell. And I had such a fear I was doing the same thing at StartPlaying. I was simply running to the next thing the moment things got hard.

But here’s an exercise that helps. Take that thought. Examine it. Pick it up. Take it seriously. Put it on a pedestal. Give it armor and a steed.

Are things hard?

No. Things are easier than they’ve ever been. We’re running hypotheses. We’re testing. We’re researching. We’re building based on data. The team is working harder and faster than they’ve ever been. Each week we’re hitting our best week ever. Politics are nearly non-existent.

Am I running?

I’m leaving. Yes. But not in a hurry. I spent a full month sitting on this feeling, letting it ride.

And after that, did I run?

No. I did meta-learning. I talked to every expert I could think of. I had elongated conversations with ChatGPT. I read every book on business transitions. I hired the world’s expert to coach me. I spent months building concordance with Devon.

Did I rush to the next thing?

No. I spent six months forcing myself not to do what I wanted to do from the very start, to jump into some sexy new AI startup.

So why couldn’t I shake that thought?

Because you can’t just deconstruct it. You can’t just oppose and win. Not in the long term. You have to construct a vision. You have to know why.

So why did I leave?

During my downtime, I wrote my autobiography. Embarrassing to say. Vain, almost. But I had a reason. When I asked folks, what did you do during your break, those precious few months between hamster wheel and treadmill? They answered: travel, meditate, rest. But why? To reflect.

I’m not a meditator. I’m not a hiker. I’m no beach bum. But I am a writer. Writing is how I reflect. And when I wrote my autobiography, I discovered something quite interesting.

Some periods were remarkably full. My high school years. My early and late college. The first four years of StartPlaying were so incredibly dense with memory. Like a thick baklava. Or moist tiramisu.

But the last year was calm. Peaceful. Surprisingly relaxing. And I barely remember a thing about it.

“Why, Nick? I left because my page per unit time was far too low. That’s it. Call me a sucker for adventure. But by the time I die, I want a thick ass autobiography.”

Devon smiled. Raised his cup. Nick and I did the same.

“To thick asses.”

We all laughed. And I knew after that drink, it was the right time to leave.