Menú

Microservices as complex adaptive systems

05 September, 2019 | 32 min 40 sec
Podcast Host Mike Mason | Podcast Guest James Lewis
Listen on these platforms

Brief Summary

James Lewis, a principal consultant at ThoughtWorks, has been exploring how to apply theories of complex adaptive systems to technology, in particular the practices and characteristics that underpin successful uses of microservices. Join our regular host Mike Mason to hear more about the surprising world of complex adaptive systems, and what technologists can learn from it.

Podcast Transcript


Mike Mason:

Hi everyone and welcome to the podcast. My name is Mike Mason and I'm here today with James Lewis. Hi James.


James Lewis:

Hi Mike. Hi everyone.


Mike Mason:

You might know James as Mr. Microservices. He is one of the original authors of a whole bunch of literature on microservices. I remember Martin Fowler stole him across the Atlantic and hid James away in his attic for a week and they wrote the original article on microservices. James talks a lot about microservices at conferences. And this topic on the podcast today, you and I have talked about it a little bit, but I don't know very much about it so I'm going in very much as a novice to this topic. But you've described it as the physics of microservices.


James Lewis:

Yeah, I appreciate that sounds a bit hyperbolic or slightly grandiose.


Mike Mason:

Just slightly.


James Lewis:

Just slightly, yeah. But I think actually, what we're going to be talking about which is this idea of complex adaptive systems, does underpin a lot of the practices, principles, characteristics that have made microservices so successful. We talk about microservices, we talk about organization and our business capabilities, we talk about long-lived product teams, we talk about smarts and the end points and not in the pipes. So, these characteristics tend to be the things that, when teams adopt them, they become successful with the microservice architectural style. And conversely, when they don't adopt these organizational design patterns, if you like, they tend not to be so successful with microservices.


James Lewis:

And I guess over the last five years, it's been five years now since we published the article, Martin and myself, which is scary. But over the five years I've seen, time has moved on and I've been doing a lot more reading and learning, and research into stuff that I guess could be the underlying organizing principles of why this stuff works.


Mike Mason:

So why do microservices work?


James Lewis:

So why do microservices work, yeah. And really, this came together for me beginning of this year when I was in a conversation at an event with someone from Amazon. And they were talking about how, as AWS gets bigger, as in, as they add more teams, it's easier for them to get bigger. So it becomes easier, the bigger they get to get bigger.


Mike Mason:

Because that's quite counter-intuitive. Most organizations would say that as they get bigger, the complexity increases and they're managing complexity, and actually for a lot of places managing legacy and things like that. Those all become anchors and inhibitors to growth. But that's not the case with Amazon, or AWS?


James Lewis:

Well, AWS in particular they were talking about. But yes, exactly that and that's why I became so intrigued by this idea, that they bring super-linearly faster than linear growth. And so I've done much more research on that over the last, I guess four or five months into this topic of complex adaptive systems, which is an underpinning, organizing framework, if you like, for the study of complexity across all sorts of domains.


James Lewis:

So, cities for example, complex adaptive systems, organizations, companies, life is a complex adaptive system, computer systems are complex adaptive systems.


Mike Mason:

Okay, but we haven't defined complex adaptive system yet. Is there a coherent short-ish definition for that, or is it just an example-based thing?


James Lewis:

Yes, there is. There is a definition. You're right. We should define our terms. So the name "complex adaptive system" has come from the Santa Fe Institute, which was created some decades ago now to study complexity. It was created by a bunch of nuclear physicists and others as a cross-disciplinary way so that ecologists, biologists, chemists, economists to study, discern this idea of complexity, complexity theory. And they defined a complex adaptive system as one which displays four characteristics. So it's composed of self-similar parts. So in the case of life, the self-similar parts were cells.


Mike Mason:

Cells, correct, okay.


James Lewis:

Those things are self-organizing. So they will organize into particular things. In the case of life, your lungs turned into lungs over time, right? As you're growing. You get this idea of emergence. So, emergent behavior, where the whole is greater than the sum of the parts. And actually, Murray Gell-Mann who passed away this week, he wrote a book called The Quark and the Jaguar, which is about this idea that how can you get this emergent complexity, emergent behavior of life, of the jaguar from something as seemingly simple as subatomic particles. And then the final thing is this idea of complexity. Again, the whole is greater than the sum of the parts.


James Lewis:

So those are the four characteristics. And with those four characteristics, they've explored, done a lot of research into the similarities between different types of complex adaptive systems. And there are some crazy similar... Would you like to hear some weird similarities?


Mike Mason:

I'd love to hear some words weird similarities. Hit me with some weird-


James Lewis:

I'll hit you with some weird things, okay.


James Lewis:

So in terms of cities, here's an interesting thing. I think it's interesting anyway. You can describe like... So Walmart in the U.S. is a scaled-up version of a news agent to all intents and purposes, which is your mom and pop store. Based on these four principles, if you look at how organizations grow over time, how they age, and there's a ton of information out there on this. You can take a look at Walmart and you can look at the number of people and the number of employees it's got, and you can look at... So a mom and pop store and it's number of employees, and you look at how much money each of these different things make, and then you can plot them on a graph. So if size versus revenue say. And it comes out with some really, really crazy results.


James Lewis:

So for example, it's a straight line. So you can draw an exact pretty-much straight line between that revenue of a news agent, mom and pop store and giant supermarket.


Mike Mason:

And stuff in between.


James Lewis:

And stuff in between, yeah. So as an organization doubles in size, it's revenue, if you like, well actually revenue doesn't double in size. Its revenue increases at about 85% of its size. So this is the thing you're talking about a minute ago. It feels like this is a weird thing for someone at AWS to say, "As we get bigger it's easier to get bigger." So if you look at the data from most, all organizations historically, within the data they've got, they've analyzed this back to the '60s, that's not the case for the vast, vast majority of organizations. The vast majority of organizations, as they double in size actually they only grow at about 85% of revenue.


Mike Mason:

Which is still worth doing.


James Lewis:

Which is still worth doing of course. [crosstalk 00:06:59]. Which is still worth doing. But the interesting thing is why. Because there's another complex adaptive system like us mammals, which has got a similar thing with scale. So as a mammal scales, so as you double the size of a human to get, I say I must double the size of a human horse-


Mike Mason:

Horse, yeah okay. A horse.


James Lewis:

Right. So you go from a human to a horse. They're composed of the same things in the way that a mom and pop store and Walmart are composed of the same things, people, essentially at the bottom doing work. Once you go from a mom and pop store, sorry, as you go from a human to a horse, they're composed of the same cells. And you double the number of cells because there's double the amount of stuff in a horse than in a human. [inaudible 00:07:40] or whatever. But their metabolic rate doesn't double, the number of calories a horse needs to intake, to feed, to provide energy to twice the number of cells is not twice the number of calories.


Mike Mason:

Is it 85% as many?


James Lewis:

75%.


Mike Mason:

Okay. All right. So I was guessing a bit too high there.


James Lewis:

But you can draw the same line. So you can draw the line from the smallest mammal, I think it's called the Etruscan shrew, up to elephants and blue whales. And as you double again, double in size, double the number of cells in the mammal's body, metabolic rate scales at 75%.


Mike Mason:

At 75%.


James Lewis:

So as you get bigger, you get more efficient, essentially. There's other weird things as well.


James Lewis:

So if you look at certain cities, as you double the size of a town, you get economies that scale for infrastructure costs to the tune of about 85% as well. Same as companies. So, double the size of the city, you have 85% of the amount of actual water pipes, or electricity cables or et cetera, et cetera, et cetera. All the infrastructure, sewage, blah, blah, blah.


Mike Mason:

Now that one makes intuitive sense to me for some reason.


James Lewis:

Right.


Mike Mason:

Okay.


James Lewis:

And the funny thing is as you say, is the why. When you've got all these data points, which plot as straight lines on graphs, there might be something-


Mike Mason:

That means something.


James Lewis:

... there, right? Exactly, yeah. And so over the last 20 years, there's been this research done at the Santa Fe Institute, actually led by... I can't remember this coauthor's name, which is embarrassing. But Geoffrey West I think was the first author. He's another nuclear physicist. And he postulated that these effects could be explained by three simple proxies, three very simple ideas. The first is invariant terminating nodes. So cells in a body, for example, people in organizations and people in cities, and houses in cities-


Mike Mason:

Invariant terminating nodes.


James Lewis:

Invariant terminating nodes.


Mike Mason:

So the the cell is always the finish line.


James Lewis:

More or less, yeah, yeah, yeah.


Mike Mason:

The bottom-most tree leaf in the tree or whatever.


James Lewis:

Yeah.


Mike Mason:

All right.


James Lewis:

The second thing is, what's above that, which is the tree. This idea of, it's called a hierarchical space-filling network, a fractal space-filling network. So in biology, in mammals, the space for the networks are essentially our blood vessels and nervous systems and things. They're essentially big trees that start off with our circulatory system at the heart, and end up in the capillaries. And those capillaries are a minimum size because they have to get blood flowing through them. And they have to be a minimal, well there's a maximum distance between them because of the-


Mike Mason:

Because they're feeding things.


James Lewis:

Because they're feeding thing.


Mike Mason:

Cells.


James Lewis:

Invariant terminating units. And the same in organizations with information flow in this case, right? Which is a weird step to take. There's less research from the organizations and companies side, but this is the theory.


James Lewis:

Also, I should point out, we shouldn't tell our staffing function about this idea of invariant terminating units, because that will be the next greatest name-


Mike Mason:

Rather than resources, we will be called invariant terminating units.


James Lewis:

Yeah, thanks Silent Running for that. So you have this idea of hierarchical, fractal space for the networks, these terminating units, and then optimization, fast feedback mechanisms, that's tuned the structures to behave in a certain way. So in terms of life, evolution has been acting as this type of feedback, as in over millions of years.


Mike Mason:

So not particularly fast.


James Lewis:

So not particularly fast, no. Not particularly fast, but over millions and hundreds millions of years, it's optimized the circulatory system. It travels to the point where it's pretty-much optimal as it could possibly be.


James Lewis:

Using an example of this, which is crazy, I think it's crazy. And again, this is from this research. So when blood flows from the aorta to further arteries say, as it passes from the aorta into the two next arteries down in this network, if those arteries didn't add up together to be exactly the same cross-sectional area as the originating artery, the aorta, you would get back pressure because of wave interference and reflection of the boundary. And that would put pressure on the heart. So the heart would in fact be beating against itself. So over time, evolution has helped our heart do the least work it possibly can, by having our circulatory system perfectly optimized to reduce back pressure on the heart. And this is why things like plaques and the buildup of cholesterol is so bad because it's essentially introduced-


Mike Mason:

It narrows your arteries and creates back pressure.


James Lewis:

It's narrowing so you get this back pressure.


Mike Mason:

Okay.


James Lewis:

Right. So that's kind of a crazy thing in mammals. And actually, if you look at... Geoffrey West and his coauthors did the maths, literally did the maths on this. By using those three properties, you can explain the basal metabolic rates of three quarters, so this three quarter scale for mammals, just based on the topology of the networks and these other two principles. It's very strange that mathematically you can show that one leads to the other.


Mike Mason:

So you take one of those properties, which is the-


James Lewis:

You got the hierarchical network-


Mike Mason:

... hierarchal network, you do some maths and you come up with 75% which is the same as the observed, increase in-


James Lewis:

Yes.


Mike Mason:

... calories required because you don't... Wow.


James Lewis:

Yeah, I know. Bizarre. He's got this, a comment. I don't know exactly how he describes it, but he talks about when you see that there appears to be this similarity in things across lots of different systems. Then it sort of points to the idea that there might be something more fundamental, as you said earlier. And this is what they've found with this research, is the fundamental things that these fractal space for networks, invariant terminating units and feedback mechanisms.


James Lewis:

So what's fun about organizations is they have the same thing, right? If you think about most organizations, they have hierarchy, which is this fractal space running a hierarchical network. They have the people doing work, they have information flow and work flow through all these different structures. And it looks like, if you can do the maths, it looks like that explains this revenue growth that's sub-linear, that's, you don't quite double the 85% based on the information flowing through a hierarchical system.


Mike Mason:

Okay, I get that maybe for an information organization, but Walmart sells things.


James Lewis:

Yeah.


Mike Mason:

But you've just given me kind of an information flow reasoning.


James Lewis:

It's information but it could also be physical atoms, physical flows of work. Let's... They're moving furniture upstairs, whoever's outside. The physical, it's not so much what's flowing, it's the depth of the hierarchy that matters in terms of how long things take to happen.


James Lewis:

And so for example, the blood pressure of an elephant is the same as the blood pressure of a human, which is the same, in the case within ours, the blood pressure of a shrew. They have invariant blood pressure across all sizes. So what actually changes is heart rate. So the elephant's heart beats way, way, way slower. And in fact, an elephant's cells live at a much slower pace. In fact, elephants suffer from cancer much, much less than humans do because their metabolic rate is so much-


Mike Mason:

The cells have divided less over their life.


James Lewis:

Yes. Exactly. The only difference, literally, the only difference between it, and I often, when I'm asked, is the number of layers in the hierarchy, in the circulatory system. So an elephant has, whatever it is, 30 layers, 16 layers, something like that, of arteries, then it hits capillaries. The max is like two. And those two effects appear to correlate with this thing of metabolic rates decreasing to economies of scale. And similarly, in the organizations with an organizational hierarchy, as you add more hierarchy, not only does organizational metabolic rate slow down, if you like, so things take longer in bigger organizations-


Mike Mason:

With more hierarchy.


James Lewis:

With more hierarchy.


Mike Mason:

I'm sure none of us have observed that ever.


James Lewis:

None of us have ever observed that, no. So things take longer. But it's not just an odd thing where things seem to take ages around here. It's like physics, right? It is like a thing based on these maths and the bar, which is quite fun.


James Lewis:

But not only that, so things take longer, but essentially, the same set of properties can predict the aging effects that we see in mammals as well, and in organizations.


Mike Mason:

Wait a minute, what aging effects in organizations? Like legacy software aging effects?


James Lewis:

Well this is, so I didn't realize this but companies die way more often than you think, if you define death has been taken over or actually going out of business, whatever. Apparently the churn on the U.S. exchange is, the half-life of a company is 10 and a half years. So you get 50% churn-


Mike Mason:

And that's getting shorter over time. [crosstalk 00:16:34]


James Lewis:

Yeah, that's getting shorter. Yeah, yeah, yeah. So this idea of corporate aging and corporate death, it also leads to these ideas of complex data systems and these hierarchical networks.


James Lewis:

So there you go. I've just told you some really crazy stuff. The blood pressure of mice is the same as the blood pressure of a blue whale, more or less. We've all got about one and a half billion heartbeats in us over a lifetime, same as an elephant, same as a mouse, there are variant. That's been known for quite a while. And Walmart is the same as a mom and pop store, just scaled up in some way.


James Lewis:

But, and now here's the fun thing. Now coming back to-


Mike Mason:

Are we getting to microservices eventually?


James Lewis:

We're going to come back to microservices right now.


Mike Mason:

Oh right now, okay.


James Lewis:

Right now. You can cut as much of that out as you'd like to.


James Lewis:

We'll come back to Amazon right now. So this is, if you remember the thing I said at the start, which was that AWS appeared to be scaling with an external of 1.15, so they're growing faster than they should. Rather than it being 85%, they're getting 115%.


Mike Mason:

So if AWS doubles the number of what, staff working there-


James Lewis:

Yes.


Mike Mason:

So not servers or anything like that. So staff employed by AWS, double that, you get 115% of the what, revenue?


James Lewis:

Revenue. Yeah, yeah, yeah.


Mike Mason:

Okay.


James Lewis:

Right. So actually, what the quote was, was, "As we get bigger, it's easier to add new people." As we get bigger, it's easier to add more products or create new products. And your observation, rightly, it was actually most companies, it's the opposite. As you get bigger, it becomes harder. And that's because of the hierarchy. So this is, as you get bigger, you add more layers in the hierarchy, and as you add more layers in the hierarchy, things slow down. Things have further to travel, whether that's physical goods or whatever, that's information.


James Lewis:

So what's different about AWS? Well, there's another fun thing that cities don't just display this 85% economy of scale for infrastructure. They also display a 115% return to scale for socioeconomic effects. And this is why cities are so successful essentially. So as you double the size of the city, you have to pay 85% of the cost for things like transport, and for water pipes and all this stuff.


Mike Mason:

Infrastructure, right.


James Lewis:

Infrastructure. But for number of patents, innovation, number of professionals, things like number of universities, all these things, they scale super-linearly. They scale at a hundred-


Mike Mason:

For all activity-style measures.


James Lewis:

Yes, especially economic measures, but also crime and disease and pollution. They all scale super-linearly at the same exponent, which is 115%-


Mike Mason:

1.15.


James Lewis:

... of 1.15.


James Lewis:

So, this is why cities are so successful. The theory goes, that as you double the size you spend less on maintaining them, but you get more "productivity" out. So you get more for less, essentially. But it's the same [inaudible 00:19:30], and it looks like Amazon in a way, AWS is kind of a weird thing, there, all right.


James Lewis:

Now it turns out, Jeff Bezos is a patron of the Santa Fe Institute. And I haven't been able to dig out when he became a patron, but I've got a horrible feeling that he's done this deliberately. He's designed Amazon, AWS, in such a way as to get returns to scale. As to get this 115%. Because there is another type of network which is still fractal space-filling like the hierarchical one, but there's another type of network called a small-world social network which displays the 115% return to scale.


Mike Mason:

How old is that model then, the small-world social network?


James Lewis:

The small-world social network, the small-world network has been around for ages.


Mike Mason:

It's kind of academic theory [crosstalk 00:20:16].


James Lewis:

It is academic theory. Absolutely, yeah. We've had it for ages and it's basically based on the social interactions between people. So the theory is, based on Dunbar's numbers, this is the number of connections that we can hold with different types in our heads.


Mike Mason:

The number of humans you can know effectively and [inaudible 00:20:31].


James Lewis:

Exactly. So if you're in a city, you get to choose who those connections are. So, we're developers, right? We get to choose to go to these meetups. And that's accelerated the pace of learning and knowledge and understanding, and sharing of information in these more cohesive social groups leads to this 115% increase in innovation and biomass measured by patents and this kind of stuff, and wage growth actually and various other things.

James Lewis:

But also, crime. So that sort of stuff clumps together as well, right? So you get similar effects. But it's all based on this idea of social, Dunbar's number and the sociability of humans and how we interact with one another. But if you think, through the village, there's only 150 people in the village and you can only have 150 acquaintances or whatever, those acquaintances are just all the people in the village, including the [inaudible 00:21:18], the-


Mike Mason:

You don't have any choice about it.


James Lewis:

You don't have any choice about it. But you have the choice of who to associate within the city. So this idea of creating these small-world interlinked networks, which is what cities are essentially, these networks of individuals, that also turns out to be a complex adaptive system and it also turns out to give this 115% return to scale.


James Lewis:

And Jeff Bezos, if you go back, what 2003-4, he sent that memo, where he basically put constraints on the organization and says, "You need to be in small teams. You can only communicate by external analyzable service interfaces." Yada, yada, yada. Those constraints, you could look at them as essentially designing in to the organization a small-world network, but deliberately creating the type of network which gives you returns to scale and not just economies of scale. So this is where you start to... You know, it's not paranoia if they are out to get you, Mike, right?


Mike Mason:

I don't know, people have lots of different views on Amazon, right? So I think we're trying to stay away from that right at the moment and just pick apart what have they really managed to achieve. But it's interesting, because Amazon is very frequently in the tech world used as this kind of exemplar of what you should aim for. Like people took about two-pizza teams and frankly some of the easier stuff to comprehend about what Amazon's done. And I guess what you're saying a little bit here is the two-pizza teams is not really the crux of what they're doing. It's more about this small... Was it small-world social network?


James Lewis:

Small-world networks.


Mike Mason:

Small-world networks. And some of the activities like two-pizza teams are just an outcome of that or-


James Lewis:

There are constraints to create this. So there are constraints that's being used to... Actually, ThoughtWorks did the same. When we were opening offices originally we had the rule-



Mike Mason:

It was 150 people or something like that. I think it was-


James Lewis:

It was.


Mike Mason:

... the Dunbar number, right?


James Lewis:

Yeah. It's a similar, it's the same reason why we'd want to create offices of that certain size. You create a business, it's a maximum of that certain size. And it appears that they've done it by adding these constraints, and again, the externalized ability and saying, "You can't talk to one another anymore. The only way you talk to one another is through these interfaces."


James Lewis:

The guy from Amazon was saying, "The only time we're really synchronized is for a funding ride, which we do a number of times a year."


James Lewis:

But the teams are so decoupled from one another that they've created this ability to just add new things without any impact on the existing things that are already there. Whereas in a hierarchy, when you have new stuff, you've got to shuffle lots of stuff at the bottom of the hierarchy. You've got to add new layers of middle management in. You've got to decrease or increase the length of time it takes for information to flow, or physical stuff to flow in those hierarchies. But with this other type of network you don't get that.


James Lewis:

And this is the thing with microservices, why I think it's interesting is, apparently, we saw those characteristics made sense and the companies that were using those characteristics, the ones Martin and I described, they were being successful with them. And it appears that all those characteristics are again, pointers towards this idea of creating a smaller network. The business capabilities, product teams building those business capabilities independently from other teams and so on. So, decoupling across these teams.


Mike Mason:

Forgive me for asking a question I should know the answer to, but when you guys were writing that original microservices article, you were documenting rather than inventing.


James Lewis:

Yes.


Mike Mason:

And that's fair to say, right?


James Lewis:

Absolutely.


Mike Mason:

So you were documenting something that was working out in the world for organizations, and you reverse-engineered the characteristics that you thought were the most important ones, and they coincide with the small-world network theory.


James Lewis:

That's exactly it. Yeah. We didn't invent anything. We have to say that. It's all... If you look at things like, if you look at an organization like the Guardian, Spotify and some of the investment banks, and Amazon and Netflix at the time, and a bunch of our own internal projects, those characteristics were, as you say, extracted from what they were doing, that we thought, and they thought were making them successful. But if it appears that there's just these three things that are actually underneath, underpinning it all, you can explain all the rest of it while the rest of it is successful, just by having these three fundamental patterns.


Mike Mason:

And now we've gotten back to microservices. Tell me the three fundamental properties again because I've always... It was-


James Lewis:

Self-determination... Sorry, invariant terminating units.


Mike Mason:

So, resources, right?


James Lewis:

Resources, yeah. Hierarchical... Well, space-filling fractal networks, hierarchical or small-world. All right, so there's two different types of space-filling fractal network. One type of network implies economies of scale. So 85, [crosstalk 00:26:11] 75%. The other type of network implies returns to scale. And what microservices architecture does is it tries to guide you towards the small-world network, and we're on the internet, right? More like the worldwide web rather than creating many architectures.


Mike Mason:

Right, so the upshot of all of this, the quite interesting conversation we've just had, the very interesting conversation. I'll upgrade it to very interesting.


James Lewis:

One of the more bonkers ones on the podcast actually.


Mike Mason:

It's a good bonkers conversation though. The short version is, there's a whole bunch of very interesting stuff happening around research into complex adaptive systems, and it's just another justification for the microservices work and your useful architectural style.


James Lewis:

I was say justification, I'd say... Sorry. I'd say it's kind of, it's explanatory towards it. It's an explanation as to why it does work when it works, when it's appropriate.


Mike Mason:

Right. And is it also a clue as to why when people try to do microservices but they get it wrong somehow, does it help explain how they've gotten it wrong? Or is it not that?


James Lewis:

I think it does. And I haven't done a lot of work on this, but my suspicion would be that what you're trying to do with, if you get one of these catastrophic distributed monolith-type problems, which is one of the common failure rates we see with microservice architectures, that's potentially because you're trying to impose the wrong style of network, if you like; the hierarchical type of network-thinking on design, which is fundamentally a small-world network. And when you do treat it as these varied business capabilities, completely independently developed by product teams with multiple product teams within them, only interfacing via, on the wire essentially, via service, API to RESTful interactions. Ideally, that's essentially one of these smaller networks that you're trying to build where you've got the people, are also in this small-world network, and the way they're communicating with one another, within their local teams.


James Lewis:

So it's like a grand unified theory of organizational design and all microservice physics [crosstalk 00:28:13].


Mike Mason:

Because you were about to get into Conway's law there as well, right?


James Lewis:

I was. And the Mythical Man-Month, right? To me you could argue the Mythical Man-Month, the idea that adding more people to a failing project makes it-


Mike Mason:

Fail faster.


James Lewis:

... fail faster, yeah. When you can... Right at the start we should have said about this, you mentioned, Mike, about the idea of the capillaries can only feed a certain number of cells, they can only be a certain distance apart because if they were further apart the cells in the middle would die. It's kind of the same with, almost like adding more terminating units to the point where the flow of information stops, it can't actually disseminate it across the people, across the invariant determinates units.


James Lewis:

Conway's law I think is an expression of this as well. Conway's law is describing, if you're in a hierarchy, you think you're going to build a hierarchy, even if you're in one of these other types of networks, then potentially you can build one of these other types of networks.


James Lewis:

And you could argue that ThoughtWorks is an expression of that. Our organizational structure, you could argue that Amazon or AWS is... and actually also Valve, when they leaked their employee handbooks a number of years ago. Fascinatingly on the second page, they actually have our organizational structure, and it's a dot and then a line with all the other people underneath it, completely flat. And then underneath that they've got a smaller network diagram. This is how it actually works. We're all connected to all these different individuals. People have been doing this for awhile, we've just not been recognizing it. Maybe we're just catching up.


Mike Mason:

Awesome. Well this has been hugely fascinating James. Where can people find out more if they'd like to dig into this?


James Lewis:

Yeah, absolutely. So, the book, which I heartily recommend, the de facto reference for this stuff would be a book called Scale by Geoffrey West. It came out a few years ago. It's pretty amazing reading and it's got all this research in it. And obviously there's a ton of papers behind this as well. The original research which you can dig out in the references subsection of that, which are also fascinating reading. Not least the last chapter where he points to the other papers, which seemed to show that if cities and society keeps scaling at 115%, then some time in the next 30 years we're going to hit a finite time singularity at which points there will be a phase change in the configuration of human society according to the maths. Now, very technical, but it doesn't sound good.


Mike Mason:

How soon is that?


James Lewis:

2050s. Something [crosstalk 00:30:40]. Some of the research is pointing-


Mike Mason:

It sounded a tiny bit Ghostbusters a [crosstalk 00:30:44] at that point. Or [crosstalk 00:30:45]. I don't know.


James Lewis:

Yeah.


Mike Mason:

Okay.


James Lewis:

But this is what I think a lot of environmentalists are talking about. John Kay is talking about the idea that we can't just scale GDP exponentially. Actually, we're not even scaling GDP. We're scaling it super-exponentially in the case of cities. So, that's an issue. It's a real issue that the only way to get out of is to out-innovate it. So the only thing that stops us hitting this cliff edge, if you like, is we would need another revolution along the lines of the computer age, or the invention of the steam engine. Otherwise, head for the hills in about 2049.


Mike Mason:

Head for the hills. Join your local prepping group.


James Lewis:

Well, it's actually one of these, I mean we'll finish up now. It's one of these things, is actually it's... This civil idea is almost like the recipe to make more money than you could possibly ever imagine, which is why I suspect there's a lot of entrepreneurs, should we say, to use a nice title. We'd jump on this and can organize their companies as well, make all the money in the world. But if you do that, then you're also hastening, massively hastening, the onset of whatever weird crisis comes. So maybe we should all just work in hierarchies and shut up.


Mike Mason:

Well, we'll leave that to listeners to decide.


Mike Mason:

Okay, my name's Mike Mason. I've been here with James Lewis. Thank you very much, James.


James Lewis:

You're more than welcome, Mike. Thanks for having me.


Mike Mason:

And thanks for tuning in. Please do leave us a rating and a comment on iTunes. It does help other people find the podcast, and we'll see you next time.


Zhamak Dehghani:

Hey everyone. I'm Zhamak Dehghani. Please join us for our next episode of ThoughtWorks Podcast. We'll be talking about the role of CIO and CTOs in today's organizations.

Check out the latest edition of the Technology Radar