You Don't Get to Create Anything

Oh, Randy, my friend, it is wonderful to have you on Still Burning. We've known each other for a long time. A long time. OK, enough said about that. When did you first know that you were a geek by this definition?

Oh, wow. That's interesting because my father was a geek by this definition. He worked at Xerox PARC and was an early computer scientist. He inspired an interest in math and computers for me. But actually, when I went to university, I was planning to be an international lawyer. In high school I'd done debate and learned the political issues of the day. I graduated high school in 1986 and college in 1990, so high Cold War. What I thought I would do with my life — my dad, famous computer scientist, had lost great stuff — I wasn't scared away as much as I thought my talents were elsewhere. So long story short, I ended up getting a political science degree from Stanford, but I really love math and computers from my dad. We didn't have minors, so I ended up double majoring in what's now the data science major — mathematical and computational science. Applied math, statistics, operations research, computer science. But that was the side gig. That was for fun. Political science and international relations was the mainline career. I studied overseas in West Berlin. There was still a wall around it. Studied behind the iron curtain. So I was into the East-West relations part.

I'd had some internships during college at Intel doing software engineering. It was a great way to make money and I enjoyed it. But again, not my career. Fun, really fun. I enjoyed it. Not what I thought was my career. And then after I graduated from college, I spent two years working at Oracle as a software engineer. And then, much to the chagrin and surprise of all my buddies there, I said, OK, I'm going to take the GRE, take the LSAT, and go do my international law stuff. I started the first year of a four-year combined JD-MBA program — Stanford Law School and the School of Advanced International Studies, Johns Hopkins international relations school in DC. Started the first year of that four-year program, still on the track. In the summer between the law school and the IR school, I took a summer associateship — basically an internship for lawyers. It was on Sand Hill Road, doing IP work, patent prosecution. Getting patents for a geek was the most mind-numbing and boring and terrible job.

You don't get to create anything.

You don't get to create anything. All you do is write down the cool creation that the inventor did. So the inventors up there on the whiteboard described this wonderful, cool invention. And I'm like vibrating with excitement. Oh, did you think about this? Did you think about that? And they're like, no, no, no. You don't do that. You don't get to do that. You just write it down. And then you do bureaucratic infighting with the patent office to get the patent. And like when I had just come off of this immediate gratification — at the end of the day, you fix the bug, you've added the feature, you've done all this stuff. And I was having so much fun there. Anyway, long story short, that was a crisis of identity for me that summer. As an antidote to the boring patent law, I was reading M. Mitchell Waldrop's book, Complexity. And it was only by immersing myself in these super geeky things that I could kind of get through this summer. And so by the end of it, it was very clear what I liked. I knew what I needed to do. But it took me the whole summer to get myself ready to say to everybody — hey, everybody, the thing I've been saying I was going to do for 10 years, I'm not. And so I went back to Oracle and continued working as a software engineer. I think that is when I realized I was a true geek.

That's the highly creative part that you didn't have available to you. And then you realize, no, I can't live without this.

I never would have phrased it that way, but absolutely true. Creative in the true sense of creating a thing.

You have some mental image. Do some work. There's a thing. We can make thoughts manifest.

Yes. It's like Prometheus. Yeah.

And do you find that background — you've had a number of executive positions — do you find that background in politics and international relations playing into that?

A little bit. I don't have a counterfactual. But I will say that learning about history and politics and international relations, there are strategic elements to it. There's understanding that human relations are an important thing. But I think a thing that the social sciences brings that we don't get enough of in technical disciplines is a true appreciation for nuance. There are things that are for sure wrong in the world. A lot right now. But most of the world is gray. And whereas a lot of people gravitate toward technical disciplines because there's a right and wrong — that is true at the bits and bytes level, but it is not true at the human level. And nuance, understanding that there are multiple ways of achieving the same goal and that we do it collectively.

And things bite back sometimes.

Sometimes things bite back. And having an understanding of the history of those places, having some of the languages — I still have some German from when I was in Berlin, always kept the Spanish up — that interest in other countries, histories and politics and economics, it just makes you a fuller human too.

I read a lot of history, still do. I think it provides perspective. We have a disaster on our hands? No, no, this is not a disaster. There are some bad consequences, but any number of really horrible things have happened in the last hundred years that make this not even a bump. So you can take a situation seriously without panicking.

Right. Yeah. In the current climate, I kind of wish I hadn't taken all the classes on authoritarianism and know the playbook. But that's for another podcast.

We talked a little bit earlier about this idea that for years there's been a playbook. And those scalability techniques were part of that playbook. People have gotten really used to just being able to turn to some page in the playbook — you have a problem, there's the page, you apply the technique, and away you go. But you were there as part of the team writing the playbook. That's a very different set of skills. And one of the things I've noted about the introduction of augmented development techniques and tools is a lot of the playbook got wiped clean. And there are people who are just kind of panicked because they've never not had a playbook before. And I realized, OK, well, you can write the next one. But it's a different set of skills. And I see people copying. Oh, well, use TDD — one test, make it pass. Is that really the best thing? I don't know. It's a thing. But is it the best thing given the tools we have now? Nobody knows. Did you get bored once there was a playbook?

Oh, no. It's just as hard to execute a known playbook as it is to come up with it. And I'm not doing this just to — you're the guy I'm talking to — but like Extreme Programming Explained, 1999. I posted myself on LinkedIn — hey, I just found this new book that teaches you how to do AI. It's called Extreme Programming Explained. And subsequently, the continuous delivery book, Accelerate, now Frictionless, team topologies — there's a whole trajectory of very closely related ideas, slightly reformulated and applied to different things. But hey, it is hard. As an engineering leader to go in and say, y'all have heard about this playbook and you're doing about 1% of it. Let's do the other 99%. And the reason I know you know this, but I'm demonstrating I do — the reason why you had to write that playbook, Kent, is because it's not obvious. It is not. It's obvious to you. OK, and you're weird in the most wonderful way. But it is not intuitive that we shouldn't batch all our stuff up and really hard and like — all these iterative ideas, they're correct. They are part of the physics of developing software, but at least demonstrably to most people, they're not obvious. Breaking things down into small units, checking your work all the time — why would I write the test of a thing before I write the thing? That's crazy pants. And once you start to see what those techniques can do for you, they're revelatory.

So I get really bored once there's a playbook. I am not the least bit interested. I'm ready to move on to the next thing. So I find it interesting that you like both.

Oh, wow. OK. That's great. For me, I like to create the new thing, but I am also a more deductive thinker. The way my mind works — state the principles and then here is how to apply them. And that's the way my mind works. Like give me a platonic ideal and then I can apply it in the real world. And that is exciting to me. I don't get bored by executing somebody's playbook, especially yours. I like doing good work. The application of a playbook is a joy to see it come to fruition. I love to cook. It's only chemistry. But it's pretty complicated chemistry. I like to do new things, but also it gives me joy to do the same things I've done before and really nail it.

What are you doing with augmented development? With the genie?

Until this past week, I was the SVP of engineering for Thrive Online Grocery, about a hundred engineers. And I hired a buddy of mine, Mike Winslow — Comcast for a long time, then Amazon Music — and he's leading the team around AI augmentation across the entire engineering team and the entire company. He's running dojos — out of Target, American Airlines, there's a bunch of people who've done dojos for DevOps-type things. Dojos for AI. Bring non-engineer humans and pair them with an engineered human and build a thing together. And those have been fabulously successful. As we record this, we've just been doing it for a couple of weeks and they've been like, everybody's so excited. The legal team is excited about stuff they built. The merchandising team that buys the food, they're excited. And separately in terms of AI augmentation of software — it's all about context and bounding the genie. How can we keep the genie within the bounds that we would like? Partly that's spec-driven in the forward direction, partly that's eval and adversarial stuff in the backward direction. And we're doing both. These guys named their knowledge graph work "genome" — the idea is to come up with a collective description, to the best of our knowledge, of what does the product actually do. It's 12 years old and we have 12-year-old software and lots of people don't even know what those areas do. Connected with how the software actually behaves. So a knowledge base we can use for context for spec stuff, and separately working on a harness for eval to make sure things are correct.

So are you falling into the side project trap? Where it's just so easy to start stuff?

I don't personally. Here's where maybe I have to take away my geek credentials — I am a Luddite in my personal life. I came very late to having cell phones. When I'm at work, I love playing at the edge of the industry-level distributed systems stuff. And then when I'm at home, I'm not technical at all. So many friends are like, oh, I've got my agent to do this. And I'm like, great, good for you. What I am at home is — going to the farmer's market, cooking stuff, hanging out with my son. I have tried to get myself to be excited about technical side projects. And I just can't. It's enough for me that that's my day job and more than half my waking hours. But at work, absolutely. This is Jevons paradox. Now that essentially cognition is near free, there are all these things we would do with cognition that weren't economically viable and now they are. Jevons was an English economist in the 1850s, 1860s, talking about coal. Oh, we made the extraction of coal cheaper. Does that mean we spent less on coal? No, we actually in net spent more on coal because now coal at one pound instead of 10 pounds is now useful for all these other things we never had before.

What's a tool that you would like to have that you've never been able to build?

Wish I were more creative. I will say that this is a tool that Mike just built — a really good look at a team's execution, not to blame them but to help them: where are the bottlenecks? Rate of feature production, rate of bug fixes. You look at this sort of dashboard of the whole overall engineering organization or team by team, and it becomes so evocative to see how, oh, like it takes five days to do that thing. Well, wow, let's work on that. I'm a very visual person. If there were a way to visualize all the bottlenecks — show me the problems and show me in a visual way. That's a business model already for somebody to go make. Yeah, for sure.

And the admonition to limit WIP is one I've been saying a lot in my career, but I have to keep saying it a lot. Oh, I'm working on these five things. How about work on these two things, get them done, and you can work on the other three. My buddy Martin Thompson likes to say — when you show me you can use one thread, I'll give you another. And he's got a high bar for that.

So, last question. What about this wildly changing situation we find ourselves in — you can interpret that as the world, technology, whatever you'd like. What about that scares you? What wakes you up at night?

Oh, geez. Do you want the whole list or just the top 150? I am cautiously optimistic about the future of our country, our society, our planet — but every one of them is under threat in ways we've never seen before. And that's a big old challenge. I don't know — as humans, we've demonstrated that the only way forward is through. And what I mean by that is: to the extent that there will still be democracy, modern civilization, and a habitable planet, it is because we have used technology to do it. What if we just did solar? What if we just did renewable? It does not need to be pie in the sky. And I am confident in the long run, from Jevons paradox, that AI will create more jobs — there will still be not just the same number of engineers, but more. But with every revolution comes major disruption of the actual humans right there. In 1900, 80% of the United States population was doing agriculture. Now it is 2 to 3%. If you went to somebody in 1900 and said, hey, 78% of you all are not going to be doing farming in 126 years — that person would have said, what a dystopia. Like it wouldn't have been 100% wrong. But seriously, we've gone so far in ways that person could not have imagined. What keeps me up at night is that we don't buckle down and solve these things. We don't make it possible for ourselves to solve these things. And from the technologist and political scientist in me — we have learned over and over again that you can't shame people into doing the right stuff. Making people feel bad about hurting the environment is not going to make them save the environment. Instead, you have to make it easy and good and better for everybody. Like if we were going to electrify everything, solarize everything — it has to come with: this is objectively better for all of us. The only economic forces you can really harness are those forces of — I need life to be better for me and my kids.

Well, I'll let you work on that. What a pleasure to have you here. I really appreciate it. I appreciate your friendship.

Thanks, man. I love you, man. Thank you. Love you too.