That we don’t need to innovate in every part of what we do. At Embark, we absolutely do want to start from first principles when it comes to our primary job, which is designing software for self-driving trucks. But for things like comp structure or legal docs or PR, it’s usually better to defer to the experts, at least at first. When we decided to produce videos, for example, we initially thought we could do it ourselves. Then we realized there was absolutely no comparison between what we thought was good and what actual professionals would make. We didn’t know you needed a helicopter to shoot a truck driving down a highway.

I think it can still be valuable to start by giving something like that a shot yourself, not because you think you can do it on your own, but just to better understand the process. We did make a couple of simple videos before we hired pros. But we also use our network to connect with people who already know the best practices, and generally, we follow their advice for a while. I’m under no illusion that the default is always best. But if you don’t understand why a standard exists, you’re probably not in a great position yet to improve the state of the art.

Have a plan that’s credible to you, and then follow it. I say “credible” because you do have to be reasonably confident it could work—you need enough experience to understand at least in theory what steps to take. When my co-founder and I first built a self-driving golf cart in my garage, I’d done competitive robotics and programming for years. We could say, “We know what sensor to use there; we know roughly how you’d process that; we know where to get those parts.” If you can close your eyes and picture yourself walking through each step, there’s a good chance you know what you’re doing.

Following your plan is harder than it sounds, because there will be a heck of a lot of people who think it won’t work. When we started out, experts in robotics told us, “You’re not ready.” It was dispiriting. But the thing that kept us going was knowing that each individual part of the plan should work. We weren’t positive, of course, but we couldn’t see any reason it wouldn’t. Just because something’s never been done before doesn’t mean it can’t be done.

I stopped using Safari and a bunch of other apps on my phone, and my life is so much better. I used to go on Facebook or Hacker News whenever I had some downtime, and I felt like it owned me as much as I owned it. So I deleted everything but email, messaging apps and photos, and I had a friend lock me out of Safari with a passcode. Now I only use my phone for a specific reason, for a few minutes, and when I’m done I put it away. Plus when I’m sitting in an Uber with nothing to do, I answer a few emails, and those little chunks of time add up.

I’d like to learn Mandarin. China is the up-and-coming place right now, and I’d like to be able to connect with people there effectively. My dad grew up in a lot of different places and speaks half a dozen languages; I’ve always admired that. He’s not fluent in all of them, but it was cool to travel with him and see how he showed people the respect of learning at least some of their language. He always made sure the first thing we learned when we visited a new country was how to say, “Thank you.” It’s a good reminder for me that you can start small, because when I learn something new my default is to read everything I possibly can. I like to understand the architecture and work from first principles. But as I have less time for those deep dives, I’m getting better at just doing something—anything. It might not be perfect, but it’s enough to build momentum.

High Output Management, by Andrew Grove, which is an excellent book for new managers. One takeaway that’s helped me is the difference between general aptitude and task-specific knowledge, and how important it is to invest in people’s task-specific knowledge before you send them off on their own or evaluate their performance. Rather than just giving someone a project, for example, you can ask them to write a project proposal, then look at it together and pick it apart before they move to the next step.

I initially thought people wouldn’t be open to that approach because it’s kind of micromanage-y. But I’ve realized that while people don’t like micromanagement when they already know what they’re doing, they actually really like it when they’re not sure yet. For new employees in particular, taking the time to give context and put in gates where you can affirm they’re headed in the right direction—that builds their confidence quickly. It also helps me as a manager, because I can quickly spot opportunities for them to improve. And when they do start working on their own, I know they’re getting things right. I used to think of myself as a low-touch manager, but I’ve realized it’s more about modulating the level of touch based on where someone is in their growth cycle.

I also have Deep Work, by Cal Newport, which I wanted to read because it’s hard as a CEO to find long chunks of time where I can truly get lost in a project. And I’m reading Fire & Blood, George R.R. Martin’s new book. It’s much different from the main Game of Thrones series; it’s like a history textbook. But that’s what I like about the universe Martin created—it’s big enough to feel real.

Until recently, I’d always thought camping was super hard. I had these vivid memories from when I was a kid of struggling for hours to put up big, complicated tents and of the rain coming in. Then some friends and I went camping this summer and it turned out to be unbelievably straightforward. I did do a lot of prep work—I set up my tent in my bedroom first to make sure I could put it together—but it was great. Maybe the equipment advanced a lot or maybe my memory just failed me, but I was like, “Wow. We should do this all the time!”

I’ll say 25 minutes, the length of a pomodoro. I don’t follow that technique super strictly, but what’s useful for me is completely disconnecting from asynchronous communication for that short period, because no one should be upset that you didn’t respond for 25 minutes. At the end of that time, I’ll often just turn off Airplane Mode, check in and, if everything’s good, go right back to work rather than taking the five-minute break. But even if I do have to respond to something, I’ve still had enough time to get into a flow.

I also find time in the shower incredibly useful, because there’s nothing to do but think freely. The challenge is trying to keep track of your thoughts without a notepad. I use a technique that associates items with locations—those two projects on the corner of the wall, this one on the shelf. I can get a list of 10 or 12 things going that way.