Tesla and SpaceX with Joe Justice
Join Murray Robinson and Shane Gibson as they chat with talk to Joe Justice about how Tesla and Space X have developed a new operating model based on self-organizing teams that continuously discover, deliver, and improve to achieve a thousand year goal.
It’s like the most radical combination of open space agility, OKR’s, Continuous Discovery, Continuous Delivery and Continuous Improvement, you can think of.
Whatever you think of Elon Musk, these companies are way ahead of most other companies and accelerating further. Tune in for a fascinating discussion about how Tesla and Space X work.
Read along you will
Shane: Welcome to the No Nonsense Agile podcast. I’m Shane Gibson.
Murray: And I’m Murray Robinson.
Joe: And I’m Joe Justice. Thanks for having me.
Murray: Hi Joe. It’s great to have you on. I’ve watched a few videos where you were talking about what it was like to work at tesla, which is fascinating and bizarre, and we wanna really get into that. But why don’t we kick off and get you to tell people who you are and what your background and experience is, and how’d you get to this point in your life?
Joe: Okay. I’m from a tiny farming town and I had a dream of seeing the world. I studied as a software developer, and I went to work for a tiny rural software development company, filling out government form software. I was one of the first developers to use the predecessor to what became microsoft.net and I ended up writing the pension Benefit Guarantee E-filing software stack. So when a company says, let’s work with the United States government to guarantee retirement funds, how will companies send us the required documents? E-filing? And I did that at a fairly early age.
No one wanted the project. It’s not like I was a genius. I, lucked into this hard project I. Learned a lot, read books. I had to talk to a lot of people because most of the books on that new software stack hadn’t been published yet. So I got to directly talk to many of the people that helped create what is now.net directly. I went to the software developer conference at the time and we came in early and left after, filled in pages and pages of notebooks, talking to many of the architects, developers of what is now.net to understand how to use the thing.
And that happened to be done as one of the first Scrum projects. And this is when Scrum didn’t even necessarily have a product owner that was added later.
So I was involved in an early agile project that happened to be using what was just becoming Scrum. And using a software stack that eventually became huge. C Sharp was heavily marketed, the developer tools became quite good and a lot of people learned it. That got me a consulting job
and that did take me out to travel the world, but I was still really young and had no idea what I was doing. I had a fair understanding of dot.net, but I wasn’t the master of it. And in consulting you parachute in to some company where they have a problem and they want help and they’re paying usually your company a lot of money. So the company that is hiring you indirectly has all these high hopes for you and you arrive and I’m a kid, and I got my laptop and I’m like, okay, what do we do? And that was the next 15 years of my life. I saw hotel rooms and conference rooms all over the world. I did have some successes. I had plenty of failures. I did meet a lot of people. I did learn a lot. But that travel schedule was brutal. So I’d been in conference rooms and hotel rooms all over the world, but not seen any of the sites. I’d mostly just been trying to write software. And the huge break for me came when I was recruited by the Bill and Melinda Gates Foundation. The technical part of running these grants for philanthropic work wanted to use Agile and they wanted to use the T net stack.
I came out and had a set of meetings directly with Bill which was unreal. And Bill was just stepping back from Microsoft. And so for this little window of time, Bill had a lot of time and wanted to mentor. And At that point in Bill’s career, Bill started every single meeting by saying, how much of this can we do in parallel. Every single time. Not sequential, not what step 1, 2, 3, 4, but how can we cut. Whatever the project is, whatever the goal is, into parallel work streams. And then what Bill would do is run all of them at once and you would have these masses of parallel teams of which I got to be the agile person to try to figure out what agile would mean in a useful way to these teams. Not that I was always successful, that was my job to try to do that and I did and I learned a tremendous amount.
About that time I became involved in environmentally friendly cars. I became so passionate about that I decided to found my own car company wiki speed and go through a huge amount of paperwork to become a registered road legal automotive manufacturer.
Now there’s a thing called a kit car, which I think is completely awesome and it’s a much lighter legal load. You say, we’re gonna buy parts from these certain critical areas and then we won’t take responsibility for the final assembly and safety. The consumer has to do that, and then they would go to one of the areas in the world where you’re allowed to build your own car and the final owner operator would claim, I’ve made this vehicle safe enough for me.
But if you are going to take accountability to make a car for someone else to be a manufacturer, it is a massive body of legal documents and liability law. And I was so excited about this project. I waded through all of that and became a registered road legal automotive manufacturer which was extremely unusual because of the legal load.
And I did that by just trying to follow Bill Gates’ advice. The way Bill thought about any project seemed to be, how can we cut this not into sequential pieces, but into pieces that don’t have to wait on each other and can advance and even finish in any order. So these teams can just run as fast as they want with their relatively small piece. And that changed my life. I’d been exposed to parallel processing before, but I hadn’t experienced someone applying that to every aspect of business every time and to phenomenal effect.
Murray: That assumes that you’ve got a lot of people and resources to help though, doesn’t it? So if you’ve got a project that you’re breaking up into 10 parts with 10 streams, then that’s a lot of people at the same time so you can finish your critical path earlier.
Joe: In 2006, I published what later became called Justice’s Law. And it’s what you just said, Marie. It’s not helpful if you have more parallel executing chunks than you have units of execution which is typically a team of three to five people.
If you have 50 people, so that’s approximately 10 teams you don’t need 11 parallel executing architectural chunks and you’re suboptimized if you have nine. So there’s the, you wanna balance, and that’s half of what became called justice’s law.
The other is any company structure that doesn’t match your parallel executing units only slows you down. So if you have nine parallel executing units and nine teams of three, four or five people. If you then have a separate QA team by definition, they’re a bottleneck for everyone. You’ve just introduced a company-wide bottleneck. So that’s the final piece of Justice’s Law, which is: The modules of your products define the structure of your company. And what’s implied by that is you want to balance and you wanna avoid any other piece of company structure that’s not one of your modular pieces.
Murray: That’s the opposite of Conway’s law Conway says your product architecture is going to be a copy of your company communication structure, because that’s just the way you do things.
Joe: Melvin Conway was genius.
Murray: Yeah. So you’ve turned it the other way around and say, if you want to have a product that’s made up of these modules, then reorganize your company. So it looks like the architecture, which is gonna be best for you.
Joe: I was forced into that a consulting career. I was really thrown into these things and didn’t know what I was doing. And people would say, how should we restructure the company because you’re the agile person. And I had no idea, I’m some kid that just got a plane ticket to whatever the nearest major airport was with my laptop, and I didn’t know what was going on. And I would have to say I don’t know, I’ll get back to you. But after being asked that more than a dozen times, I’d started to develop some theories by watching what the companies did do. And you pay really close attention when the company asked you and then, together they try something, so it was able to learn deeply because of intensity. And that is how. Somebody, even like me, was able to develop an inverse to Conway’s law and then even add to that a little bit because I had so many chances. I had so many opportunities by being asked how do we restructure our company in an agile way.
Murray: Yeah. So I read about Wiki speed, a few years ago and I remember you talking about designing and building cars in one week iterations, like scrum for manufacturing. How did it go? What happened there?
Joe: Wow. Okay. An early lesson, just because you wanna finish something in one week iterations doesn’t mean you can. it did go really fast, so it was a success story. But just ’cause I say we have a one week increment of work doesn’t mean we’re I finish a car in a week. Especially in the beginning. It took 12 weeks. I was off by a factor of 12 to get something that could drive with a license plate on it legally.
Murray: And how many cars were you making at the same time?
Joe: I was absolutely resource constrained. In the beginning it was just one. The engine was on an engine stand being developed as its own module. And then over there the seats are being developed along with, seat mounts and sliders and the bottom of the frame over there. And those are being developed in parallel. And then the roll bar systems with an interface to the body is being developed over there. And so there’s these parallel pieces, but that’s definitely not a car, that didn’t emerge until later, about 12 weeks later. At peak there may have been 10 cars in development at once, and I say May because it became a global project. People would make pieces in the Canary Islands and upload video clips to YouTube saying, here’s the car pieces I just made. And I never met these people. So , it was a decentralized project. The facility that I operated, my Wiki speed shop often had four cars in development at once.
Murray: So different groups around the world could be part of this and make their own cars using the same approach and learn from each other. Is that the idea?
Joe: That’s part of the wiki speed model that I still don’t think has been evangelized or publicized or maybe even not even repeated very much, is how decentralized and frugal, like how absolutely little money was needed is maybe stupendous, maybe surprising. And how little leadership or management or central control basically none was needed. Those parts I don’t think have been explained very well. I don’t think the, those are as hot topics. People are much more interested in the modularity of it, the parallel execution of it, which is why I started with those pieces.
Murray: It reminds me of a company called Local Motors that was building the Rally Fighter. Are you familiar with them?
Joe: I went to local motors a whole bunch of times, a massive respect for their c e o and what they were doing. Ultimately, the local motors did decide to close their doors and sell off their assets. But not until after they made the Ollie self-driving three D printed minibus. Local motors really did some avant guard stuff very early. Their Ali Autonomous three D printed passenger bus, that actually was used. It had some airport loops it ran and some neighborhood loops it ran. That’s still what companies like B y D are trying to do. I have massive respect for local motors, but they did ultimately decide to pursue other things. Close their doors, sell off their assets but yeah, I actually worked with them a number of times.
Murray: Because they had that kind of wiki model, as well.
Joe: We had a lot in common. I was absolutely inspired by what the Rally fighter was and the cars they produced afterwards, and not only cars they went pretty broad, but they did do a lot of cars.
Murray: yeah. they did a lot of interesting things, so what would you say were the main lessons you learned from Wiki? Speed
Joe: You can pretty much do anything with no money. As insane as that sounds. I think it’s impossible for a company to run out of money unless they’ve already run out of motivation, in which case money is being very inefficiently used anyway. Because when you’re unmotivated, you use money to compensate for lack of motivation, and that’s an extremely inefficient use of funds.
To hop ahead a moment, that’s a reason why I believe all of the Musk companies are as successful as they are. For example, the Tesla model Y is now the bestselling car in the world.
It’s outsold the tow to Corolla. It’s taken the number one spot globally. And I think a reason that’s happened is the level of motivation is so high that the cost to achieve milestones, like the most efficient heat pump ever made or the highest gain antenna. There’s companies working just on antennas. Why is that a SpaceX product? The reason so many best in the world modules emerge from these companies is because the level of motivation is ridiculously high. So people can spend money extremely efficiently. The money is essentially just used on cost of materials.
Murray: So did you go straight from Wiki speed to Tesla ?
Joe: I had a stop in the middle at Amazon and I became a developer at Microsoft. And then John Deere, the agricultural machinery company they were the first ones to attempt to implement the Wikispeed model, and they did. It was a tremendously financially successful move. That was a bizarre story for me to learn about. That got me to work in defense. John Deere recruits a lot of military veterans from all branches of the United States Armed forces and from all levels. People who really understand how to work, even when you’re tired. Which is also true of leadership to a point. And I got to work with Raytheon, Northrop Grumman, Lockheed Martin. All of the world’s largest defense contractors on the U Ss, a front side. And that’s what brought me to Australia. And I worked with all of the branches of the Australian military and then joined Tesla and created the Agile program at Tesla.
And while I was a Tesla employee, I chose to visit SpaceX very often because I had a lot of access. I was not a SpaceX employee, but I really wanted to know how these companies were similar and how they were different. So I went to SpaceX quite often.
Murray: Let’s talk about Tesla. We hear all sorts of things from the outside about Tesla and SpaceX and Elon Musk. My impression is that those companies are hyper innovative, super fast, super efficient, but also extremely demanding, very anti-union. That’s what we hear from outside. So I’m very interested to know what it was like for you to work for Tesla.
Joe: So first the job postings talk about things like being a software developer. Being a manufacturing line engineer, once you’re hired, you have no role limit or job description. So software developers help unload trucks and weld. Welders help write software, and also unload trucks. Everyone helps unload trucks. A role description really doesn’t separate people in a Musk company the way I see it separate people in many other types of companies it’s very much a all hands on deck. So maybe that’s my first piece of a, what is it like to work in the company?
But then I’ll answer no. I was not a software developer, although I did get involved in writing some software. I was in a manufacturing role. I was working with robots and painting systems.
I ended up working or taking notes on every position in Tesla, Fremont which Toyota had that same building before. There was a General Motors factory, a huge factory in California right by the city inside the Fashionable Bay area. That was a General Motors plant and then Toyota took it over
Murray: This is the Numi plant?
Joe: That’s The Numi plant. That’s the building. And then Tesla bought it. Same building really different experience. When Toyota or GM operated that building, what they did was final assembly. So engines that are made somewhere else are brought in. Transmissions that are made somewhere else are brought in. When Musk got it. That was Musk’s one big building.
So product design is in that building, which is not how it was under Toyota or GM software development is in that one big building, which is not how it was under Toyota or GM making the motors, making the batteries. Chemical refinement for the battery chemistry is in that same building, which is not how it was under Toyota or gm.
Whether that was intentional or just lucky it was awesome because the level of cross-functionality and the level of vertical integration was higher than GM or Toyota have exhibited ever before or since. And that is why so many best in the world pieces of engineering come out of there.
It’s more possible to understand the holistic product process; design to build when it all happens in one building than when you have some of the software designed in Hungary and some of the software built in India, and that’s just software. And then the metal too, like the plastics for the headlights are from a company in Thailand, but they’re made by a company in China and that’s just the plastic on the headlight, and that’s how it is in most companies.
In the Musk companies, it’s not. The group of people that is choosing which plastic and then melting it to test it for the headlight lenses are three meters away from the people who are writing the software that’s gonna auto adjust the headlights. And that makes these synergies, that makes the rate of innovation able to be different.
Murray: And they must be right next to where people are installing those lights into the cars. So if they see that they don’t work, they just change it straight away.
Joe: Yes. And those same people are gonna end up also installing them. The idea that you’re a software developer, but you’re gonna help with whatever needs to be done. You are a manufacturing line robotics engineer, but you’re also going to help do whatever needs to be done. That means you’re not just near where it’s happening, probably more than once a day. You are doing that other job because of this all hands on deck mentality.
Shane: It is like DevOps, but for car manufacturing in terms of you design it, you build it, you fit it, you fix it. Was that a conscious decision, do you think? So that vertical integration was, that was the conscious choice, vertical integration. Therefore buy one factory, put a series of constraints, and the constraints will drive and dictate the way you work. Or was it a case of, we bought a factory, everybody was in the factory, and that constraint then forced the vertical integration?
Joe: Shane, that’s exactly my big question. Unfortunately, I’ve only got the question. I don’t have the answer. I became a Tesla employee in 2020. My guess is Musk wanted to completely vertically integrate, was really happy to just have one building, but also it was budget constrained, likely.
My guess is everyone under Musk who at that time came from traditional manufacturing, there was a lot of people from Mercedes-Benz. And you can really see that in the Model Ss. It’s so similar to a Mercedes-Benz product of that time. It’s radically different now. And I work with Mercedes-Benz now. Those cars are completely different now, but at that time they were extremely similar. So you had a lot of people from Mercedes-Benz who I would guess really wanted separate buildings. They wanted their beautiful office far away from the noise and dirt and the feeling of class privilege of having their different parking lot and different cafeteria, et cetera. I imagine and Musk, I imagine really wanted vertical integration in one building. So I imagine that if the budget had been higher at the time, there would’ve been a compromise reached. Because Musk had some really powerful personalities then from Mercedes-Benz mostly who probably would’ve been able to dictate some level of compromise. And I think we wouldn’t have had the level of vertical integration. But, the company was cash constrained. There was just one building, they backed into it.
Shane: So just to follow on from that, we talked about Bill Gates and paralyzation of multiple teams. And then we talked about wiki speed in terms of complete decentralization, and the hive mind to build something and then onto Tesla with, constraints based vertical integration with those other patterns. How did alignment happen, so I can imagine with vertical integration and a constraints based model in that building, you’re seeing what’s been built, you’re getting constant feedback. If the fenders don’t fit, you go whack the fender right and redesign it with the parallel model, we’ve got nine teams running. If there’s any interdependencies, we’re gonna get outta sync. We’re not gonna build anything. And decentralized car making, same thing, the people around the world can build a part that just doesn’t fit with what everybody else is building. So how the hell did you get, that collaboration across decentralized teams.
Joe: Awesome point. For wiki speed. It actually ended up being easy. I was working alone in this little garage in Denver, Colorado. I created the first set of eight interfaces. I didn’t know what the modules were gonna be, but I created the interfaces and I did a lot of basic engineering calculations. How strong do these need to be and in what dimensions? Does it need to be really strong against twisting force or lateral load or which direction does it need to be strong and how strong? Okay, what’s an interface that I can afford to make, which I had no budget that can handle that force. And then electrical and data too. Do I need to know what data flows across? Let’s make sure I at least have an ethernet cable there so I can figure out what data I need later, am I gonna need a lot of electricity? Okay, well let’s put positive and a negative plug. That’s a big enough wire diameter to handle some power, let’s figure that out. So I did that.
And then as people did join, Wow, that was lucky. I would say you can make anything you want as long as it solves these problems. There was a problems list for each module, solve these problems, please. Can anyone figure it out? ’cause I don’t know how, and it does need to meet these interfaces or it’s not done. Whatever cool thing you came up with, if it doesn’t meet these interfaces, I don’t know how to use it, no matter how cool it is. So it’s not done yet. And that was frustrating for some people that joined because they had a very interesting idea, but they hadn’t thought about bringing it to meet the interface that someone else had designed.
Murray: That’s a lot like community coding projects, with GitHub where they have these long lists of problems to be solved and I’m gonna be working on this next, and somebody comes in and after they pass some validation, they get to pick something up.
Joe: I think it might be exactly the same. I have never joined a community development project. But I think it’s super similar. There’s a documentary about, a Finnish group of kids who are now highly successful tech adults. But at that time they were kids and they were in what’s called the demo scene. Using the smallest amount of computer memory as you can to make the most interesting visual and sound performance you can. And the group is called Second Reality. And they shocked the world. These kids, legitimate children out of Finland in no space at all. Like some of their programs were 64 kilobits. But the one I’m gonna talk about is a little bit bigger. Second reality. It was in Dolby surround sound when people had sound blaster audio cards. And this was in Dolby surround.
And it had the set of sequences of these different, very impressive visual performances. One of them is a polygon fly-through of a spaceship through a town that looks like it’s out of star fox or X-wing from a decade later. It is unreal. And then these kids are doing it. , part of why is because the code is so efficient that they could run it on the machines of the time. Like it’s just really tight. Code. It’s the real deal. I actually think that was my inspiration for what Wikispeed was because I had never joined a community development project. That, and what I thought I understood of the Wikipedia author model, even though I had not contributed as an author or an editor at that time.
Murray: So you’re at Tesla, you’ve got this software development background, but you’re actually working with robots that are doing painting and you are coding them and you’re fixing them on the assembly line, and then all of a sudden you’re unloading trucks. How do you know what you should be doing?
Joe: So work was almost never alone. We were almost always at Tesla in a group of about 50 people.. That then self-organized into groups of three to five people, occasionally, two. And we took turns who was physically touching the keyboard or taking the actuator apart or buffing the paint on the car.
Murray: So this is the pair programming mobbing pattern, right?
Joe: We didn’t call it any of those things, but I think it’s super similar to mobbing. Mobbing is absolutely the most similar thing that I’ve seen to it. And because of that, I was very often doing something I’d never done.
Murray: yeah. This is what people naturally do when there’s an incident, like an outage in your company. Three or four people will just stop everything and gather around the keyboard and say, what’s happened? What can I do? Somebody might lead it, it could be anybody. It could be the junior person is leading it, but then people go away and they draw stuff on boards . it’s kind of a natural response to a serious issue. But you are talking about doing it all the time, aren’t you?
Joe: There was not really anything else. A, useful side effect of doing it all the time is even if one person there had done that type of work. All the time and had deep expertise. There was always at least one person there who had no idea what it is we were about to do. What that means is at least once a day, everyone was faced with something they’d never seen before.
Even if it was routine work. What that means is when there was something that no one had ever seen before, we knew how to handle it. It was the same way we handled everything else. It was a work style that was not surprised by the first time you’ve ever encountered something, and that made encountering something completely new possible, maybe even fun.
Shane: And I’m gonna assume that there was no concept of a methodology. You turned up you had the supposed role, you got the fright of your life to realize that actually the role was just a piece of paper to get in the door, and now you’re just part of the team and the team’s got a hive mind and you got no idea what you’re turning up to tomorrow. You just got absorbed by the organism and the organism helped you figure out where you fitted.
Joe: Yes, that was exactly what I wanted to communicate, and with fewer words than I’d been using.
Shane: Yeah. Self-organizing in a true sense not pretending. Yeah.
Murray: I watched some of your talks and you said you had an app and that it had all of the important things that needed to be done and somebody was prioritizing them. And I don’t know how that happened. And then you were saying, our team is gonna grab this problem. Can you run through all of that? How did you know what you should be working on?
Joe: Let me start with Team Wiki speed and then build a bridge to what I experienced at Tesla. So in Team Wiki speed, when it was just me, I needed some way to organize my thinking. A whole car was super tough. Once I had introduced, I. Some level of modularity. It was a lot easier. I could think of one module at a time and I’ve made a lot more progress, but I started putting post-it notes on the wall of the garage I was working in to try to organize what was the next experiment I could guess to try and what was the result of the last experiment on each of the modules.
And I started putting those on YouTube. that was great. Some people ended up giving me good advice, but, so I had a, free storage of what the last experiment was and the result so I could go back and look. Then once people started coming, those post-it notes on the wall became really useful because people would come in without any idea, without any context, just interested, and they could see there’s modules and they could see what the modules were, what the last experiment was on each module, and then what my guess of the next one was.
And they could then write new experiments or try my suggested experiment or whatever they could self-organize around the work. Wiki Speed tried a few tools. We tried Jira for a while. We tried Trello, scrummy Cara for a while, but they were all digitalizations of columns of post-it notes, one for each module.
Then I walk into Tesla and to me it looked the same. It was just far more advanced in AI assisted and with much more polished software, but it looked like the same idea. You have this board of initiatives that we guess are important. Here’s what we guess is the useful thing to try. There’s a button on your phone where you can add a new mission or category if you have a good idea and just walk to the one you think you can help. That’s how Wiki Speed ran, that’s how Open space conferences ran. That didn’t feel that weird. What was really cool is how elegantly simple it had become, because that code, was available to everyone. So anyone could propose a check-in, and say, I think I’ve made it a little simpler, a little easier to read. Or, I deleted some legacy code that wasn’t being used very much. People could contribute pretty easily.
People were really motivated. The level of motivation was really high. So having access is enough ’cause the people were motivated so they would want to make an improvement. And then all of the data of everything Tesla does is available to be processed by anyone’s machine learning experiment.
And there are so many good machine learning libraries because of the autopilot team and the Tesla vision team that everyone has access to that people would experiment by taking different types of data and piping it through these best in the world machine learning engines.
And the end result is this to-do list, this board of what we could work on would get suggestions from the AI and from people. And all these things that you wish would happen worked reasonably well. This was 2020. From what I hear, talking to some of the people I still keep in contact with. Three years on is it’s way better now. At that time it didn’t have a name. Now it does have a name. It’s called Tesla One. It looks phenomenally elegant from what I can tell. In 2020, it was a bunch of very different looking screens. It was a lot of different apps and you’d switch between apps to try to answer different types of questions.
If I was in the paint area, like I said, I did spend a lot of time in paint. The software looked different than when I was over by Stamp and Die, even though they used a lot of the same software libraries behind the scenes. Now it seems they’re unified under Tesla one.
Murray: Who set the priorities and the objectives and the goals? How were they set? Because it, I can understand that a small group of people can pick things up off a board, but was management doing that or were the goals, even the big goals also set at the low level.
Joe: okay. There is no king that I could find in Tesla, not even Elon. Elon was really involved, but involved in the engineering, not in decision making. There was no traditional management. It’s not a thing. And now most of us are so experienced in work with the idea of leadership and management that it’s hard to imagine a different system, but that doesn’t mean there aren’t other ways. So how does it work?
How does it work? Musk did set a goal for each company. The Tesla goal is accelerating the transition to sustainable energy. For SpaceX, it’s spreading the light of consciousness out among the stars. So Musk chooses those goals and that I think most of us can understand whoever’s paying for it chooses the purpose of the company. And Musk did. So the idea is the goal should last each company more than a thousand years. So that thousand year goal then has KPIs. And as far as I can tell, those are not set by any boardroom. I never found a boardroom anywhere. They’re not set by any management committee. It looks like they’re thumbs up, thumbed down Reddit style by all employees, as far as I can tell. That’s what it looked like. It looks like you turn Reddit loose basically on that thousand year goal to say what’s a number that is useful for now towards that goal.
For example: so for SpaceX spread the light of consciousness out among the stars. Okay, cool. How do we start making habitats airy rebreathers? What do we do? How do you decide? What’s the bottlenecking measure, the limiting measure to that happening? I. And Elon, I think, did propose this one, but I believe it was unanimously voted up, like it just made sense. And it’s cost per kilogram into low earth orbit. Here’s the rationale. Even if we made the best air rebreather, but we couldn’t afford to get it to Mars, doesn’t matter.
Doesn’t matter how good the things we have are if we can’t get the things out among the stars. So that is the limiting factor. And once stated just about everyone seemed to agree, and I think that was actually brilliant. So knock down the cost per kilogram to low earth orbit. So now you have a number, and now you have enormous amount of engineering freedom.
What’s your bright idea? In a group of 3, 4, 5 people aided by AI tools, you have authority to spend money. You have all these robots and stuff. Try to improve that number and you can come up with sub KPIs to contribute to that. You’ve gotta see the link. You’ve gotta make the link like how long can this material last at this operating temperature because that means we have longer use cycles in this part of the rocket engine.
Murray: Is everybody proposing these sub KPIs and they’re getting voted up through these apps.
Joe: Yeah. Through apps on their phone. Yeah. Which is really similar to an open space conference. So imagine open space conference with hundreds of millions of dollars and the biggest rocket humans have ever made, and you kind of have SpaceX and Tesla.
Murray: a lot of trust you’re putting in people.
Joe: The motivation level seems really important. And this is the part of Elon’s personality. I. And I know as soon as we start talking about Elon’s personality because of media reports, people think they know Elon, like people don’t seem to think they know the personality of the current c e o of McDonald’s, but people seem to think they know the personality of Elon Musk, even though what they do is read news reports and news reports tend to have some level of agenda. So it’s weird, but I’m gonna try to enter this dangerous conversation about Elon’s personality anyway because I think it’s where the motivation comes from.
Murray: I’ve, noticed that organizations have a culture, and a personality and a structure which is a reflection of the strengths and weaknesses of their founders and leaders. There is quite a lot about these companies that do reflect the way Musk looks at the world and does things.
Joe: That’s a primary goal of x. Now x has a lot of very interesting ambitions. A primary goal of X is to be a AI recommendation engine, which it’s not now, not publicly. It’s still basically a information sharing service with some ad revenue to support it. And now even like a truthiness score attempt through community notes. The goal of it is to be an AI recommendation service that essentially is what Elon Musk wishes Elon could be, and X as a holding company for the other Musk companies. It’s basically to become an AI that runs those companies in the place of Elon. So that’s its thousand year goal is to keep those companies on what would arguably be asserted to be a worthwhile track for a thousand years and not have that disappear whenever Elon becomes incompetent or incapable or passes away.
Shane: So just go back to this theory of set of vision and set motivation. There’s a pattern sitting under the covers. So we talked about the Tesla factory being constrained. Therefore, vertical integration was mandatory. And Elon has multiple, very large companies with lots of people. So he’s naturally brought in a constraint based model where he just doesn’t have the time to sit at the top table and answer any fricking questions because all he is gonna do is he is gonna come in. Be motivated, be quite energetic and then get the hell outta dodge ’cause he is gotta go do it to the next one. He has actually got a constraints based model on himself that stops him being the autocratic leader that we typically see with other large tech companies.
Joe: Shane, I have never thought about that way. I think you’re exactly right. Yeah. Elon definitely doesn’t have the time to be the king and sit at any board table, but that would likely imply that Elon doesn’t want a board table because that means someone has a seat that Elon might not be able to influence.
So that might be why there is no management structure the way most of us are accustomed to looking. And instead there’s a bunch of apps with transparent rules, which is what I experienced replacing management in Tesla and what it looked like when I visited SpaceX.
Interesting. What you just said is likely the time constraint has prevented Elon from being the king of the castle. And then what I’ll add to that is that’s probably why Elon doesn’t want a boardroom, because that means there’s a board when Elon’s not there. So what would you use as your proxy? The best case if you trusted software, which Elon does and I do too, is you would have those decisions made by software as transparently as possible.
Shane: And then it gives you a different lens to Twitter , because he went into an organization and implemented the constraints based model. Now it was brutal. And implementing a constraints based model after an organization’s formed and has a way of working is a very violent and volatile way of doing it compared to doing it from the beginning. So just taking those patterns And applying them to an organization that already exists.
Murray: Can I segue into asking you about intensity? Because I’m hearing intensity quite a bit and I’m wondering what that means. It sounds like a core value.
Joe: If you watch kids play Lego, sometimes they’re really casual about it. . Sometimes they get really intense, and sometimes that even makes people fight and argue, . But sometimes the pace of building works really well and they build super fast, whatever it is they wanted to build.
That analogy is really useful to me. When I experience Tesla, when I look at my journal entries, I actually took a lot of video while I was in Tesla, but I don’t release any of that. I haven’t gotten permission to share that publicly. But when I view it again for myself, like when I check myself before I make a tweet on an X looking at my video and reading my journals, that really fits that intensity. It was super high. But the reason that I use the analogy of kids playing Lego is that’s the kind of intensity, my experience of it was a lot like playing Lego.
Murray: Now there is a dark side to intensity and that is the working hours. I remember you saying you used to get there at 4:00 AM and worked till 7:00 PM
Joe: very often.
Murray: Why on Earth would anybody do that?
Joe: Okay. Let me try to frame this as honestly as I can. First is I did choose to retire from Tesla and that was the primary reason is I wanted more time to do other things, including recover. Because you work three and intense days, and you get four days off and the next week you have four days on, three days off. But it was one of those four days off and I was like, I got four days off. I’m gonna go hiking a national park. And I did, which was, intense. So people are intense anyway, by choice.
Murray: Was it like 24 hour, overlapping shifts happening there?
Joe: Yeah. Yeah. So the requirement for all work is you’re on no later than 5:00 AM or 5:00 PM and you’re there until the next 5:00 AM or pm so you can do a handoff because no work stops. No software, no production, no security, no maintenance. Nothing stops ever. So that means you are either the 5:00 AM to 5:00 PM or more, or you’re the 5:00 PM to 5:00 AM Now it’s in a group and you have this, very supporting feeling. And it’s tremendously fun. I never had more fun at work. Okay. I actually had as much fun in Wiki speed and actually wiki speed and Tesla to me were very similar. But outside of that, I’d never experienced anything as fun, but it was intense. So you need downtime. I think you need recovery time. Like someone who loves a sport and plays it often, you don’t only play the sport, you also sleep and ice your joints and do other things, right?
You need that time. That’s part of playing the sport is that self-maintenance time, that recovery time. And a mistake I made at least one four day weekend. I didn’t give myself any downtime and that just about physically broke me and I was and am in really good shape. And that was a big decision in my want to retire. So I understand how burnout can happen. I also understand how, some people have been with Tesla since founding and are still there, and SpaceX too. So it looks like it is humanly possible to enjoy that work for more than a decade.
And I think just how some people play sport and sometimes with some intensity, sometimes you just have to have recovery. You have to know how to take care of yourself and like I said, you get four days off at Tesla one week and three days off the next. So you can, if you prioritize self-care and are probably clever about how you use yourself.
I think it works sustainably. And it’s super fun and I don’t think it’s unique to Tesla or SpaceX. I actually think anyone who plays competitive sport is doing that or more.
Shane: So it’s the same as if you’re a founder though. If you’re a founder of a startup, you are burning the hours typically, you are working continuously for years until you get success or you fail. So was there that underlying feeling that everybody working for the company was like a founder? ’cause we’re using the term self-organizing, we’re using a whole lot of behaviors that sound like founders in a startup behavior to me.
Joe: This is the point. Every other aspect that we’ve talked about I think wouldn’t be enough if it weren’t for this. It isn’t quite the same as Founder’s Mindset, but it’s not far off. The way I understand it’s very similar.
I remember getting an email from Elon and this one was addressed to all staff . It was a 48.9 degree Celsius Day outside. It was 120 degrees Fahrenheit in LA County. That was insane. It had been hot all summer, and , there were raging wildfires. This was 2020. Many highways were closed because of large burning masses that had fallen across the highways. It looked like Armageddon and this is a major city. We’re in the Bay Area. We’re in a global mega city. It looks like sunset all day. It’s actually very beautiful because that’s the color of sunlight through heavy forest smoke. It’s pink, so you have pink light, and then the ash was so thick it looked like it was black snow. All day. Now it doesn’t come down. It floats side to side, but it was thick. We were told don’t go outside. It’s not safe to breathe outside. And this was the early days of Covid. We didn’t really know how covid spread. So I had latex gloves on at work. I had goggles, I had a face shield over two face masks and a Tyvek suit. There’s cooling in the factory, but it was still crazy hot. And I have multiple layers of self-protection ’cause we don’t know what these germs are or not.
And this email comes from Elon saying there’s never been a hotter recorded day in this area. We might be running out of time.
I have never been so motivated at work in my life, ever for anything. And I don’t think I was the only one who felt that way.
Murray: We might be running outta time for what? To fix the environment through by moving people to sustainable energy.
Joe: That’s just the piece. You get to pick the nuance of the, for what? The phrasing of for what? ’cause that’s all Elon said in that email. But remember the mission of Tesla, where I was an employee, was accelerate the transition to sustainable energy. So it was a work email.
So at least one likely phrasing is we might be running out of time to accelerate the transition to sustainable energy. And then you can nuance that phrase to actually have a few different agendas. We could save the world, but he didn’t say that. It’s, if this is a goal you care about accelerating the transition to sustainable energy, we might be running out of time..
That’s how Elon manages. Elon doesn’t say, look, you wanna spend money, I need to approve it or not. Elon doesn’t do that. Not that I experienced. Elon doesn’t say, you wanna get a promotion, you have to jump through these hoops. Elon doesn’t seem to do that. That was Elon’s management style as I experienced it, and it was awesome because then you look at this board of these potential projects, missions goals to try to accelerate the transition to sustainable energy and you’re like, I want to get one of these done right now. Where can I add value? It was the most fun I ever had at work.
Murray: All right. We better go to summaries, I think, Shane.
Shane: Oh, good dokey. So Some really interesting patterns came outta that one. So we talked about this idea of parallel processing, and then we actually answered the question of how do you get alignment? And the way you talked about Wiki speed interfaces makes me think of the pattern of microservices. So you’re basically saying, I have these things and they have to talk to each other. I dunno what they are. Probably don’t even know how to, they’re gonna talk to each other, but they just have to talk to each other. So if you are first you set the standard, everybody’s gonna talk to you. And if you are second, you follow that standard.
And if you don’t talk, you’re not done. So carry on and make the thing work, or it’s not useful. I have to think a bit more about this idea of defining your product architecture and then restructuring your organization to match it. I’m familiar with Conway’s Law. It kinda makes sense to me flipping the model. It’s well articulated that I just need to think about it a lot more to understand it.
And the idea is just ’cause you wanna finish in an increment of a week doesn’t mean you will takes as long as it takes. That forcing function of having a week to build a car, you end up taking 12. But that’s good you didn’t take 52.
This idea of constraints based being a conscious decision. One factory, one place means vertical integration, you have no choice. There was a role in applying for it, but once you got there, there was no rules. There were just a bunch of people with great skills and they self-organized themselves to get the job done. They did that because there was a, vision being set and there was a high level of motivation. There was a backlog of ideas, and you can pick the idea that you’d work on, but there was a form of self organization from every else to stop you going and doing something on your own. So convincing five other people to do something is a good way of validating what needs to be done.
And data was really available. That was the other key point to me, the ability to get feedback from what everybody was working on was really available and not hidden.
And then this idea that you could always see what the last experiment was. So you could always see what people had tried. When you said about wiki speed, where you wrote down what you thought you might do next.
The one about working 12 hours is really interesting. ’cause what it means is you actually work 14 because there’s always the handover period. So you’re gonna do 12 hours off of work, but you’re always gonna be an hour early to get handover and then an hour late to handover, so it’s always gonna be a 14 hour day, but that’s what you sign up for. So as long as you’re aware of that, as long as you manage your lifestyle then it can work.
So constraint based patterns, but you’ve gotta have motivation. You’ve gotta have empowerment. It sounded very familiar to Spotify and the key thing I got outta those conversations is not the Spotify model, but the Spotify culture. It always came back to strategy and culture. Strategy and culture. And what I think I got outta today was potentially the view you get of Elon from the media is may not be who he truly is, but the second one was there’s some kind of magic of setting culture mission and a bunch of talented people, and then it just works.
That’s what I got outta it. Quite a bunch of patterns underlying everything you said, which is what I like. Yeah. Cool. Murray, what do you got?
Murray: I’ve experienced unconferences and I really liked it a lot. I’ve experienced mobbing and it’s very good. But I’ve mostly worked for very traditional organizations in middle management in the engineering area or product development. And what you are talking about is so radically different. I think it’d be impossible for a traditional organization to change to this model because you’ve got a large number of people in the hierarchy whose pay, their salary, their status, their sense of identity is all tied up with how many people they’re managing and their control, their dominance, their status.
For most organizations, rising through management is about playing the political dominance game. It’s not about achieving an outcome. Innovation is something that managers just crush in normal organizations. ’cause it’s a threat, and people speaking out and saying, here’s a problem. This is a serious problem. We need to fix this. That is completely unacceptable and a lot of organizations, you can get fired for doing that. Even if it’s absolutely true. Most organizations are like Borders, they would rather ignore online until it gets to such a point that they’re being crushed only than will they do something about it.
I really liked the operating model you’ve described. I love the app. I love the Reddit style management and the mobbing approach. But I wonder if it’s possible to implement this Tesla or Space X operating model in existing organizations. It seems radical. And I doubt that management will allow it because it’ll be too threatening to their power, to their status and their income in the organization
Now. There is something that worries me about this approach and that is the possibility that people would be exploited. So when you talked about working in 120 degree day in all of that stuff, that was dangerous. People could have died or got very sick. And Elon Musk was so focused on a thousand year goal that it feels like he was not thinking about people at all. And this is probably the most common negative feedback you get when people talk about Elon and Elon’s companies. People don’t really exist for Elon. They’re chess pieces on the board to achieve the goal. And while I can understand you might be highly motivated, shouldn’t management have some sort of pastoral care for their staff?
Joe: So imagine you’re the New Zealand All Blacks or a top sports team and the coach says, I have a theory of how we need to train if we want to maintain our status as a top team, or actually become even better. It’s up to the players to choose if they’re going to continue as players and do the practice and do the training, and I don’t think you fault the coach for having the skill and making the suggestion on how to be the best team in the world.
Murray: People could push themselves so hard. They get badly injured in their training. There are people who should be responsible for health and safety in training. ’cause people don’t know, I dunno that’s my concern. You can push yourself too hard and too far. And also you might end up working so many hours that you’re actually being paid at below market rate for what you are doing and who’s the beneficiary of that the shareholders not you.
Shane: Cause you worked in little squads. I’m guessing the pastoral care was done by the people you worked with, not by a person that had the title of hr,
Murray: yeah. How did pastoral care happen?
Joe: There’s two HR groups in Tesla. And I’d say the measurements that an HR group is able to publicly use to compare companies are things like workplace accidents, and attrition. Tesla leads the other automotive manufacturers, fewer safety incidents, actually less attrition. In terms of safety incidents, Tesla has fewer than all other manufacturers. They’re the safety leader In terms of attrition, burnout is a thing, but it’s a thing in Volkswagen too. It’s a thing in Skoda and Seat and Ferrari and Lamborghini and Toyota and Mitsubishi as well, all of which I’ve worked with. The attrition rate is lower in Tesla, but Exactly, Shane, to your point, what I actually experienced is my team seemed to have a level of personal interest and check-in and safety. I experienced a very loving family feeling. It was how I experienced it. Now, I did choose to retire. I didn’t say, this is what I want to do forever. This is completely sustainable for me. Even with other interests that are passionate and energetic. It wasn’t. I wanted to do things like go hiking in the national parks and then have a job that I could relax in. A job that was less intense than going hiking. But yes, . The group care for each other. I experience that from the people around me.
Murray: All right. Let’s give people your links. How can people find you?
Joe: Joe, justice on X is the fastest response from me. I am on LinkedIn and I post on LinkedIn and I read on LinkedIn. You can reach me there, Joe Justice, Facebook as well. I use it for business now ’cause I’m often in Japan. Then I do have a website. A b i, agile Business Institute is that acronym, abi-agile.com. That has links to my book. I do courses. I’m really excited about artificial intelligence assisted group work, and I do it every day. I try to put a name on it. Please use this name as a favor to me because I’d like it to be understandable by people. I call it MOB ai, and I use that with a hashtag everywhere, hashtag mob ai. If you wanna mob AI with me, I do that online regularly. You can find me through that website, ABI Agile.
I’m doing it in person on a Europe tour, October and the first two weeks of November in Stockholm, in Finland, in Amsterdam, in Munich. You can see that list on X and Facebook, on LinkedIn, because that’s where you could do it in person with me.
Most of the time I’m in Japan, so if you’re not in Japan, we’d be doing it online, but I am doing a Europe tour. those would be the best ways to reach me.
I’ve got a question for our listeners. I actually think I understand how to create that recipe. I think I can reliably create a company with that level of focus and motivation, as huge as that is to say. But I think that’s my skill out of all this, I did learn something. And then my question is, okay, what should that company do? Who wants that power? And what do you want that power to do? Because I think I know how. Okay. What do you want to do?
You can exit me, you can tweet at me. I’m at Joe Justice on X, on Twitter. Put it in the comments to this, wherever you experience this conversation. I’d like to know what would you want to accomplish and let’s see what the really worthwhile sounding ones are and maybe we’ll make some real interesting change in the world.
Murray: All right. Well, this has been fascinating. Thanks for coming on, Joe.
Joe: Absolutely. My pleasure, Marie. Absolutely my pleasure, I hope we do have a coffee together somewhere in the world someday, but even remote, collaborating with good audio. I had a great time. I learned a lot from you both, and I hope our listeners enjoy themselves.
Murray: Yeah, I think this has been a really good one.
That was the No Nonsense Agile Podcast from Murray Robinson and Shane Gibson. If you’d like help to create high value digital products and services, contact murray at evolve. co. That’s evolve with a zero. Thanks for listening.
Subscribe to the AgileData Podcast
We will email when we publish a new episode, no spam, pinky promise