Benedict Evans — Flare Talks
Join us in a conversation with the independent analyst Benedict Evans about tech evolution, cloud, mobile, streaming, Web3, bundling and subscriptions, with a pinch of history and philosophy in the mix. The Twitter & Elon Musk saga is also discussed.
Benedict Evans's weekly newsletter analysing mobile, media and technology in general reaches more than 170,000 subscribers and his presentations and comments are well regarded in the industry.
So, hello Benedict and welcome to our Cloudflare TV segment. Hello. Hello. You're talking with us from where?
From London. You've covered this area for a number of years now.
You've seen differences between trends. And I know you also have a background in history.
So, I think you see the way technology and even the world works in a historic kind of mindset, really.
In that mindset, what were for you the main highlights in the past, I don't know, 30 years maybe, that changed a little bit the way we're living today?
What were the main highlights, you'd say?
The last 30 years. So, let's see. So, that's the 80s. Well, I mean, I think, you know, this is sort of a very basic long-term evolution of technology.
You know, start computing.
You could sort of date the prehistory of computing back to things like adding machines and typewriters and mechanical calculators at the beginning of the 20th century.
But kind of in our lifetimes, we've had sort of a couple of generational shifts.
So, we had the sort of shift to the mainframe in the 60s and 70s, which was the first sort of general purpose computing or sort of modularised general purpose computing devices.
And then we went from mainframes to PCs in the course of the 80s, although big companies didn't shift until the 90s.
And then we went from the PC to the web as the centre of the tech industry, and then from the web to smartphones.
And each of those sort of took 15 years, give or take. So, mainframes are mid-60s to late 70s and PC and then the web and then smartphone.
And with each of those, several things happen.
One of them is that everyone thinks the new thing is a toy.
All the people who work on the old thing think the new thing is a toy and it won't scale.
You can't use it for real work and it's just for hobbyists.
And for the first five years or so, they're right. But what happens is that the new thing unlocks so much greater kind of investment or finds a much bigger market, even though it maybe can't do the things that the old thing did, but it can do something less for many, many more people than would have used the old thing.
And so it pulls in all the investment in all the innovation and more creation.
And so that enables it to grow to kind of outscale the old thing. And so, you know, in 1980 or even 1975, 1980, 1985, PCs couldn't remotely, client server couldn't remotely compete with mainframes, but today no one will buy a new mainframe.
And then the same thing with the web. You know, nobody's written Windows.
You couldn't write. The idea that you could replace Photoshop or Office with a web application in 1995 would have been kind of a joke.
But nobody's actually started a company to write new PC software since about 2000.
And then the same thing with smartphones.
And it took people a long time to understand that smartphones, you know, to begin with, everyone from the early 2000s knew that mobile was going to be something.
But we all kind of thought it was going to be a PC accessory. And so we had this sort of phrase, mobile Internet.
No one really says mobile Internet anymore.
It's like saying color TV. And so what for sort of 15 years, everything was a PC accessory.
And now everything is a smartphone accessory and a PC is a smartphone accessory.
And so now for as close to five billion people on Earth who have a smartphone and about out of about five and a half billion adults.
So that's become the universal device.
And so that prompts a bunch of questions. So one of them is, well, we've got sort of 15 years in and smartphones are boring now.
Like we get it.
We understand how it all works. It's like it happened. Now what? And so that gets you conversations about Web3 and crypto and about Metaverse or VR and AR.
It also gets you questions about kind of dominant market power, which ironically are exactly the conversations people had about Microsoft.
Like nobody's been afraid.
Nobody in the tech industry has been afraid of Microsoft for 20 years. And IBM and IBM was carried on being a big company and well into the 2000s.
But no one was scared of it.
Microsoft is a vast company, but no one's scared of it anymore.
It doesn't control the agenda anymore. And so you have these conversations about market power.
And, you know, Google and Facebook, the new GM or is that the right analogy?
And then you also have sort of questions of, well, if each of the previous ones was a step change in scale, smartphones have got to everyone.
There's no other step change in scale, at least not the same kind. So that model changes.
And so what does that mean? And meanwhile, you could say there's a sort of mass deployment phase in that sort of the first 50 years of the car industry was what's a car?
What's a car company? And the second 50 years was what happens when everyone has a car?
Which gets you suburbia and all sorts of good things and all sorts of bad things.
And the same thing now with computers, you could say like from like the mid 60s or even since the war through to today, the question was, well, what's the computer?
Who's going to have one? What's the software? What's the software company?
What does that mean? Why would you have one of these things?
And now we know the answer to that, like smartphones and everyone. So then the question is, what happens when everybody has one of these things?
And so that gets you to, you know, things like Airbnb and Uber or Shein where you're not really selling software.
You're you or the Netflix and you're selling using software as a way to change some other industry, which is sort of like, say, Wal-Mart uses cars or uses the fact that everyone owns a car to change how retail works.
So you've got these kind of like these kind of ongoing sort of long cycles, long cycle innovation within tech.
Meanwhile, you've also, of course, got the sort of the other sort of trend in parallel, of course, is, you know, the emergence of databases in the 70s and 80s, the emergence of SQL and relational databases in the RP in the 80s and 90s.
The growth of open source in the 90s and the 2000s, really the 2000s. That's when I'm that's when it became kind of an industrial thing.
And now the growth of cloud and sort of the shift to cloud, which is kind of what's captured by this phrase.
Digital transformation is, you know, it's another generational change, just as you know, in the 70s, big companies got mainframes.
And in the 90s, they went to client server.
Now they go to cloud. And so there's other sort of cyclical change that's going on right now and which, again, will take 20 years for everything to move over.
So that's your 30 year question. Exactly. Technology makes things easier.
You cannot go back. Even if you were used to before, like doing the camera rolls and develop your own photography.
Now it's difficult to go back when you are used to something like that.
Right. Yeah. So I have a so I have a book from which is on Google Books from 1910 from Burroughs, which was one of the big companies making electromechanical, making mechanical adding machines.
So basically something that was kind of like a typewriter, except it's got 200, 200, 300 buttons on the front of the giant crank handle.
And that goes those things go through until the 60s and they get electrified until mainframes come along.
And it's this kind of marketing book that's kind of 200 pages long.
And it's I mean, it's hilarious because it's got this whole description of like abacuses and counting with on your fingers and things.
And then it's got the kind of creation story of Burroughs making 50 different prototypes, none of which work until he gets it working.
But then he's got all these kind of descriptions of how much time this saves and how much more efficient it is.
So, you know, he has kind of, you know, here are adding up 150 rows by hand as opposed to adding up 50 rows on an adding machine.
And you think, wait a minute. Yeah, that is what you had to do. You run the accounting department of a company in 1900.
It's literally being done, rows of people sitting at tables with a pen and a piece of paper, adding it up.
Exactly. So you kind of shift to that model, you know, and so we've got the area services, I was looking at the quote, a man will not to be employed at a task which a machine can perform.
Which is exactly the same thing. Completely. Computers shouldn't ask you to do something.
Exactly. It's like a philosophy there, right?
The philosophy that maintains throughout the years there, in a sense.
What changes if you just shift what kinds of things you realize could be automated.
And that's on the one hand with sort of things like workflow. On the other hand, it's things like machine learning, where you think, well, what is machine learning is pattern recognition.
Okay, well, what patterns? What kind of patterns?
I mean, I had this conversation with the board of the supermarket, where I said, you know, imagine it's 1977 or 78.
And somebody explains what SQL is to you. And you say, that's very interesting.
So we can now easily do queries of X by Y by Z in our data.
So like, and someone says, yeah, yeah. So you can look up your customers by their zip code.
And you say, but we don't have a custom zip code. So what would we look up?
Exactly. Now, I don't think everyone kind of understands like what FTP is and how it works and how supermarkets run on this stuff.
But it wasn't necessarily obvious when this stuff first appeared, like exactly what you could would become databases.
And the same thing with machine learning. You go to the supermarket, you say, so you can do pattern recognition.
Okay, like what? Exactly.
Initially, it's not that obvious. You don't need to recognize cat pictures. So what is it that we would do?
And we've got kind of 20 years of working out what it is that can become a pattern recognition question.
What it is that can become particularly, I think, what can become an image recognition question.
That's interesting that the image sensor becomes kind of a universal input.
And we've got we've got kind of 20 years of converting the stuff that was being done on on-prem to cloud.
And it is the stuff that's still on mainframe to cloud. And 20 years of kind of converting stuff from Excel to email into workload.
And then it will go back again.
I mean, it's like, you know, you're talking about kind of long cycle innovation.
You know, you could basically frame the whole history of technology as bundling and unbundling or as open and closed or as centralized, decentralized.
So server to client and back again. Absolutely. What cloud is, you know, you go from the mainframe to the PC, client server, whether you've got an SAP application running on the PC, connecting to a central database.
And then you go back to access.
And we'll go back again. Completely. Yeah. The evolution is there always.
I found interesting in terms of data. We're talking about databases. And I saw this article just highlighting how important it was for everyone to know about mortality rates in countries.
Something so simple like that. And it started a number of years ago, of course.
And now we go to the extreme where machine learning databases also, but machine learning is helping to find new materials to do specific things differently.
Something that humans cannot see the patterns there with so many data.
I mean, machine learning. I mean, if you think about the sort of evolution of databases, you know, the sort of the first wave.
I'm going to simplify grotesquely, but like there is a phase of moving to computers where what you're basically doing is electrifying your filing cabinets.
And, you know, if you think about a punch card that's literally stored in a mechanical, electromechanical system, it is literally you've electrified your filing cabinet.
And it's then with SQL, it goes from being a data storage system to being a sort of a system where you can ask questions.
And so you get phrases like business intelligence, like you can actually like.
And so that gets you the idea and that gets you things like just in time supply chains.
And so you don't need to have, you know, because you can process the information so much more quickly, you can actually run the business differently.
It's not just that you type into a computer instead of going around with a clipboard.
It's that once it's being processed in that way and you have the data instantly instead of once a month, once a week.
If you have the data, if it takes you a minute instead of a week to get that question.
That doesn't just like save you on clerk salaries. It actually means you run the business differently.
Exactly. Completely. It's like a different control in a sense.
Yeah. And I mean, machine learning is sort of interesting. Well, one of them.
It's kind of a cliche, but one of the kind of the dynamics maybe of machine learning is you could sort of say there's sort of two ways, two phases.
And so there's one phase, which is sort of pattern recognition in the sense that it can recognise anything that the person can recognise.
It can probably recognise anything that a 10 year old could do.
This is very things that are easy for people and hard for computers.
So, you know, the obvious thing is it's very easy for us to recognise a cat, but very hard for us to explain how you recognise a cat.
Why is that?
But practically, why is that credit card transaction weird? What is the difference between a dark shadow on the roof and a flood on the roof?
You know, it's things where you could train a 10 year old, you could get a 10 year old to do that.
You probably get a dog to do that. It just needs that kind of mammal brain ability to recognise that is a fire and not a shadow.
And so I used to describe this as basically giving you infinite interns.
So I listened to every single call coming into the call centre and tell me which ones the customer sounds anxious, which ones the call centre agent is rude.
Like you could get a 10 year old, but you can't listen to a billion phone calls.
So you basically you have infinite interns.
It's automation, just like SQL. Exactly. The other thing, which is kind of to your point and also a lot of what DeepMind is doing, is like what if you had one intern who's infinitely fast?
So what if you had one, instead of having a million interns listen to your phone calls and saying, find me all the places where the customer sounds suspicious or where the call service agent is rude.
Instead, what if you had one intern to listen to every phone call?
Because then they would start recognising patterns that the million interns could never see.
But when you're hearing a million phone calls a day, then you would go, you know what?
I just noticed this thing.
And that's some truth in that in those patterns. And there's also lessons that humanity has used throughout the years to get better at something.
You see the patterns and you act on those patterns. You know better seeing the patterns how to act.
Even for a tech company, every company, every government can look at patterns taken from that mathematical way, in a sense, and act.
The challenge, as I said, is to work out what those patterns would be. You've got kind of 20 years to do that.
I mean, just as it took us kind of 20 years to work out what to do.
We're still working on new things you can do with databases. When you go and tap your card to go into the subway, you don't say I'm using a database now.
It's just a payment card. Obviously, machine learning, each of these kind of come with their own challenges.
Databases came with all kinds of civil liberties questions, mainly because stuff that was always sort of theoretically possible suddenly becomes really easy at massive scale.
So it was kind of theoretically possible for a policeman to follow you around town.
True. But with databases, suddenly they can just like see everything you do and see all of your transactions and they can scan the whole population looking at things.
And so this is a great area of concern in the late 70s and 80s.
And the same thing with machine learning.
Like, well, theoretically, a policeman can carry a water poster around with him.
We don't have a problem with that. So what if you've got 10 million cameras all around the country and all the water posters?
What do we think about that?
Yeah. There's a lot of questions. We're not sure. And, you know, maybe it's OK, maybe it's not.
But it's not the same. Something has changed when you do that. And equally, of course, you have kind of questions of sort of machine learning bias, which is, you know, it's back to the point of this will do anything that you could train a dog to do.
But the dog isn't necessarily doing what the dog is doing, what you trained it to do.
That may not be what you think you trained it to do. Exactly.
Yeah. And there's the explanation of how machine learning and automated system got there.
If we don't know how it got there, we don't know if it's bias, how much bias it is.
So it's important to know how it got there. And again, this is not a new problem.
I mean, there's been a huge scandal in the UK where. So in the UK, the post office, most of it is a sort of independent trader.
So the post, the actual post office branches are independent retailers.
And in about 2000 or so, the post office introduced the National Computing System from Fujitsu.
And this had bugs in it that caused shortfalls in the cash balances.
And the post office saw these shortfalls and thought, wow, we've just discovered that all of our partners have been stealing from us.
And now now with this new computer system, we can see it.
And it wasn't that at all. They were just bugs in the system that were losing the money.
And it wasn't funny because hundreds of people went to prison.
People were bankrupted. People committed suicide because they were accused, falsely accused of stealing hundreds of thousands of pounds.
And this is 1970s technology.
It's not machine learning. It's just, you know, it's old technology and it's an institutional structure that was not able to understand this and was not able to ask, well, what might be going wrong?
And so the same thing with machine learning, like it creates new ways for people to screw up computers.
And the solution is not to pass a law that says you're not allowed to screw up computers.
It's to have those institutional structures to understand. No, it could be wrong.
Computer could be wrong. Exactly. Checks and balances for for those things also is also important.
Yeah, I mean, the example I love is somebody working on a skin cancer system.
And so the question is, well, what pictures have you given it?
And the obvious thing that people think of is that maybe you haven't put the right distribution of skin colors.
But the actual problem that came up was all the pictures of unhealthy skin that had a blemish on the dermatologist generally put a ruler in the photograph to give you scale.
And the picture of healthy skin didn't have rulers. So your statistical engine has been given a billion pictures labeled A and a billion pictures labeled B and asked, well, spot the difference.
All the pictures labeled A, you know that A means skin cancer, but the system doesn't know that.
It just looked at the pictures. And so all of those seem to have these transparent plastic things.
And none of those do.
See, they've built a fantastically efficient ruler recognizing system.
Exactly. Maybe all the healthy pictures are taken under incandescent light and the unhealthy pictures are taken under fluorescent light.
So you actually wouldn't be able to see that even looking at the picture, which is to the point the dog is doing what you trained it to do.
That may not may not necessarily be what you think you trained it to do.
Given that that's a very interesting topic, given that for your in your perspective, what is the most interesting trend on the Internet at the moment?
We can go either way.
But what do you think is the most interesting trend regulation or not?
Well, I would say I'm not sure I describe regulation as interesting. It's kind of all true.
It's there. But in many ways, we're sort of in a lull in that, you know, we had this kind of great locomotive pulling everything forward in the form of a smartphone.
And that was kind of the whole center of most questions in consumer tech.
And now that's happened. And so you have many different things going on. You have continuous waves of new social of companies trying to break out and be the next Snapchat.
Continuous waves of people with some kind of a gimmick that hopefully isn't enough of a gimmick that it doesn't fade away, but can become a bridge into some new behavior, which is what Snapchat did and then what TikTok did.
But Snapchat as well. You have the sort of the evolution of each consumer sector to a sort of omnichannel digital first world.
And so SHEIN is now the biggest fast fashion brand in the USA.
And it's, I think, the fourth biggest women's fashion brand in total, teenage women's fashion brand in total.
And there's no stores.
It's an app and it's a very different supply chain and a very different shipping experience and a fast ad budget.
And so you're kind of reconfiguring what it means to be a retailer, how you build a retailer, how you build a brand, whether that's in fashion or in banking or insurance or in any kind of consumer consumer, any kind of way of going to market.
You have the sort of ongoing digital transformation story, which we sort of talked about.
And then you have I mean, the way that I did this big annual presentation and I sort of the way I phrased it is you've got kind of three steps.
So the first step is you've got a lot of people in tech who are spending all their time thinking about stuff that will basically only happen in 2030, which is crypto and Web3 and metaverse and AR and VR.
Maybe 2025 if you're optimistic, but let's say 2030 realistically, because like the iPhone shipped in 2017, the smartphones didn't really become a big thing until 2007 didn't become a big thing until 2010.
So like two years away.
And it's built a lot of energy going into it, but it's not here yet.
And then you have a whole other chunk of the tech industry basically deploying ideas from like 2010, like cloud, two sided marketplaces, machine learning or machine learning 2014-15.
But you know, ideas from 35 years ago and just taking them and deploying them in new fields.
So, you know, we've just worked on how we can do a two sided marketplace in this field or that field or where we can use image recognition in this area or that area and a new way of going to market.
So these are basically ideas from the last decade and just deploying them over and over again.
And then the third point, third step is people in the rest of the economy getting screwed up by ideas from 2000 to 1995.
Like maybe people will buy online or maybe everyone will be online or maybe people will watch streaming video.
And, you know, that's what's happening to the TV industry at the moment.
It's like maybe people will watch video on the Internet.
Well, this is an idea from like 1997, maybe a long time for it to happen.
And so you've got this kind of interesting kind of disconnect between the conversations, because on the one hand, you have all these people in crypto and Web3 talking about building decentralized networks with nobody in control.
And meanwhile, the EU and the UK are passing major pieces of legislation regulating how Web 2.0 social networks work.
And you have people talking about how you might build cloud hosting on a blockchain.
And meanwhile, only sort of 80 percent of enterprise IT spending is still on Chrome.
So it hasn't even gone to the cloud yet.
Never mind, never mind blockchain. And you have, you know, people creating NFTs when, you know, most outside or, you know, a great chance of, well, this way.
So in the UK, 40 percent of non -food retail is now online. But in Italy, only about a third of people made any online purchase at all in the last three months on Eurostat data from the end of last year.
Even with the pandemic, right? Even with the pandemic.
And there's a third in the pandemic. And even then, only a third of people made an online purchase in the last three months.
When questioned, which was like late 1999, late 2021.
So the future is here, but it's unevenly distributed.
And you've got kind of some people who are living in Web3 and you've got some people who are, you know, installing the latest IBM software on their mainframe.
Yeah, they like that diversity there. But in terms of the promise of Web3, for example, and the open Internet, in a sense, of course, regulation is coming.
Europe only a few days ago approved the Digital Service Act. But I'm interested to hear you about the differences here about security and privacy versus openness and interoperability in terms of apps, in terms of services.
There's a conflict there, right, in terms of what is private and secure.
You cannot have it both ways, in a sense, in terms of the tech services, right?
So, I mean, I think it's kind of a high level. The way that I tend to think about regulation of tech is it's a bit like regulating cars.
Imagine it's 1970 and you say, well, oh, my God, look what cars have done to our cities.
And maybe this wasn't such a great idea after all.
And what do we think about this? But as soon as you say, well, we're going to regulate cars and you actually think about that for a minute, you realize we don't.
We regulate safety and emissions and speed limits and parking.
And we think about what our tax policy does for housing density.
And we wonder whether we should build more freeways or more light rail. We think we worry about whether the car company has abusive contracts with its dealers.
And that's got nothing at all to do with teenage boys getting drunk and driving too fast.
True. There are safety tests for cars even. Yes, there's safety tests for cars.
And that has absolutely nothing to do with whether you should pedestrianize Amsterdam.
Exactly. Yeah. And neither of those are antitrust questions at all.
Antitrust is not how you put airbags in cars. It's got nothing to do with it.
There's 20 different questions. And most of those questions are actually kind of complicated.
Like, you know, it took us 75 years to put seatbelts in cars and another probably 30 years to make them compulsory.
And you do airbags. And many of these things kind of come with tradeoffs.
Like, you know, you could just ban cars in London.
I think most people understand that might be a little bit simplistic.
True. Yes, I kind of understand. Yeah, that's probably. Although London, I think it has the worst traffic in the world.
We're going to need to think about that a little bit more.
And you get the kind of the same sort of thing when.
And equally, you know, people understand. But we sort of understand that.
And you also understand that if you went to Honda and said, you've got to make a car that can't crash.
Well, they could do that. They would limit the speed to three miles an hour.
And you have a man walking in front with a red flag. Exactly. All the cars have to be the same way.
Yeah. So what is it that we actually can do and what are the tradeoffs around that?
And I think in most field areas, the difference in tech is because we didn't grow up with this stuff.
We don't really have that innate understanding of what the tradeoffs are.
And so generally, when people in tech say, oh, this is complicated and we can't do that.
Well, people in every industry say it's complicated and you can't do that.
Actually, people in cars say that.
But the point is, it always is complicated. And you have to understand that tech isn't actually any different.
Tech policy isn't any easier or simpler than health care policy or education policy or transport policy or energy policy or anything else.
It's complicated and there's tradeoffs and you can't have everything you want all at the same time.
There's sort of two differences in tech, I think.
One of them is that because tech became so big so quickly, we didn't kind of grow up with it.
It took a long time to get 300 million cars in America.
It didn't take much less time to get to whatever it is, 200 million people in America using Facebook.
So because it happened so quickly, we didn't kind of grow up with it and have that innate understanding of how it works.
And the second thing is that because it happened so quickly, you can't spend five years.
You can kind of take your time to work out and understand the problem. Whereas in tech, the whole thing will have changed by the time you get there.
So by the time America passes the privacy law, half the people in America will be using a blockchain-based social network.
So what does that mean? True. And new challenges are ahead because things already changed.
And when things change, the reality changes in a sense, really.
They are. I mean, I think, as I said, to kind of repeat the point, on one level, technology policy is hard because it's policy.
It's like any other kind of policy.
You can't have everything you want all at the same time.
You can't do anything else. The difference is because it changes so quickly, you almost need kind of structurally different ways of thinking about that.
Because, you know, railways, you know, we had railways for 150 years before, well, 100 years before airliners come along.
We have airliners for 100 years before we have airliners come along.
But mainframes only got 15 years. PC got kind of 15 years of being the dominant tech.
And so if you're kind of sitting down now and thinking, how do I cut up Facebook?
Well, A, yes, you could break Facebook up, but that wouldn't actually change anything about teenage girls looking at cell phone content on Instagram.
That's not an antitrust problem.
It wouldn't even make it any easier to compete with Instagram. But B, by the time you've done that, like we might be building entirely new stuff in completely different ways.
So you might be sort of solving a problem from like generations ago, which just means like the kind of the theory and practice of tech regulation is intellectually challenging.
Not just because you have to kind of understand what an app store is, but you have to do it so much quicker and think about it in sort of structurally different ways to how you would regulate a supermarket.
Just because of that speed. In terms of the argument, of course, we, especially, for example, in the case of Apple, the app store, you were at Web Summit, me too.
Apple was there promoting, do not put another app store on the iPhone, trying to lobby Europe not to go there.
And of course, there's Europe trying to give more possibilities to Europeans.
What do you think of those trade-offs in a sense?
So I think, first of all, and it sort of somewhat irritates me that one doesn't have to say this, you have to understand that there are actually trade-offs here.
I mean, you could create, you know, Apple could make a system in which I can tap on a link in a web page and it can install an app.
That would make my iPhone much less secure because that would be a huge new attack vector for anybody who wants to put a malicious app onto my phone.
And that's a trade-off. You know, you can choose that that's a trade -off that you want to have, but you do have to acknowledge that that is a trade -off.
And you are losing something when you make it easier to install apps from untrusted sources because it creates a new opportunity.
You kind of have to think about kind of the adversarial relationship in these networks.
The same thing with, for example, making messaging apps interoperable, which was kind of a very sudden, random, last minute insertion to the DMA.
If you think of messaging as sort of like email and they're all just messages, then of course you should be able to do that.
Well, no. What happens if I want to send a Snapchat story to someone on iMessage?
What happens then? Is Apple obliged to build Snapchat stories?
Exactly. The same way. What happens if Snapchat stories work differently to WhatsApp stories?
So does everybody have to implement everybody else's features or do only the lowest common denominator features work?
How do you do encryption? That's actually an unsolved technical problem.
You have these sort of technology advocates, you know, ideological advocates going around.
Oh, it's really, really easy. Well, no, I actually talked to crypto experts.
It's like, no, it's actually not. It's very easy to do encryption within one network.
It's really hard to do it across different networks with different namespaces.
You know, what happens if I make an app that makes it really easy for me to spam as many people as possible?
Is Facebook Messenger obliged to allow me to interconnect?
Like, well, you know, you kind of need the devil is in the details.
True, yeah. And you need to credit, you know, when you look at the kind of non-public drafts of the kind of the later versions of the DMA, there's like, instead of the one paragraph that says you must allow any app to interoperate without any restrictions, which is like, no, now suddenly it's 10 pages.
Which is, yeah, OK, because it's kind of complicated. Even Facebook didn't make that as a, still they want to do that interoperability between Instagram, WhatsApp and Messenger.
Yeah, it's been three years and they haven't got it working yet.
They didn't do it. Tell me again how it's really easy. I mean, they're the owners of those apps, so.
Yeah, they can't make it work. I mean, I think there's a sort of a high level point here, which is quite a lot of these sorts.
They said there's a kind of a class of these proposals, which is things like, you know, the sort of very egregious things like, you know, Apple's 30% tax.
There's a class of stuff that everyone would sort of agree with.
And the same with the DSA. You know, you must have a rigorous content moderation policy.
Well, yeah, Facebook has one.
The bigger players. Yes, they do. The big ones have already. Yeah, exactly. So that's one side of it.
The other side is there's a lot here that is sort of from wish lists from a certain kind of ideological activist base that really wishes modern computing didn't work the way it did.
But it often doesn't really understand or accept.
Well, why is it that we got here? And it wasn't because of evil corporations.
Why is it that that's a tradeoff? And many consumers prefer that tradeoff.
And the reason there are five billion people doing it is that we actually chose to take that path rather than another path and taking the other path.
No one bought it. And so which kind of gets you to the sense, well, you know, if you were to create a model that said you are allowed to install a third party app store.
Why would anyone do that? Actually, why would you do that as a user? If you are allowed to, you know, you are allowed to change the default SMS app on your phone.
Well, never mind the security issues around that. Well, you know, a lot of these things are sort of they're things that a certain kind of person who wants to run Linux on their home computer wishes they could do on their phone.
But it doesn't follow that anyone else is going to want to do that.
So one sort of sometimes looks at these clauses and think, well, that's kind of simplistic and that's going to be a huge amount of aggravation.
Nobody's going to do that. Yeah, that's a difficulty there.
Do I want to interconnect my WhatsApp messages with my Instagram messages, with my LinkedIn messages?
Not really. Exactly. There are different platforms, different people you talk to.
So there's differences there also.
I'm also curious in the sense of about Twitter, about Elon Musk, about mainly the public town square that social media is right now.
There's a lot of discussion now because of Elon Musk and Twitter.
But of course, Twitter, isn't that a bigger, big player in terms of social media?
But it has a big influence because of those who are in the platform in a sense.
What do you think of Elon Musk buying Twitter?
It's interesting. So it's kind of cliche. I mean, I'd be packing up with this.
Twitter, there's a famous quote from Mark Zuckerberg from sort of, I think he was quoted in 2013, where he said, Twitter is such a mess.
It's like they drove a plan car to a gold mine and fell in.
And it's always been this place where there were lots of great people, but they could never get the culture and the flow to produce great product.
And it's kind of interesting to wonder, like, is the reason the product has always been kind of a mess and never really done anything great?
And it always kind of felt like it was riding the wave rather than like creating stuff.
Is that because of the dysfunction of the company or is the company dysfunctional because it's so hard to work out what Twitter should be?
If you kind of contrast it with Instagram, you know, which is a very well run company with a great product.
Does it look well run because the product actually has easier questions? It's a lot easier to explain what Instagram is than to explain what Twitter is.
Completely. The concept is very clear. Is the reason that it looks well run because it's got easier problems?
Or does the problems look easier because it's a well run company?
And the answer is probably kind of both. I think the thing now is like, OK, so now that there's somebody running it.
Oh, yeah, another part time CEO, but a slightly more accomplished one, perhaps.
Is this a turnaround?
Is there somebody going to take it by the scruff of the neck and do the Ari Gold act for the paintball entourage and turn it into a place that ships great product?
I mean, for me, like the kind of the iconic moment of Twitter in the last few years was that Substack.
So it's like I compared it to Craigslist. So like Substack launches and people said someone said Substack is Twitter's payroll.
You build your audience on Twitter and then you monetize on Substack. Actually, my newsletter isn't on Substack, but it's the same point.
You know, I promote it on Twitter.
So Twitter buys this newsletter product that had not really worked and in trying to integrate it.
And then in January, like a month, six weeks after he's no longer CEO, Jack Dorsey is on Twitter saying, why would anyone use Substack when you could use Ghost?
Like you just bought and built your own newsletter product.
Exactly. The other one. How does that work? How would you feel if you were on Twitter's newsletter team and you see the guy who's been your CEO for five years telling everybody he's a competitor?
Never mind that. How would you feel if you had decided to try and build a business on Twitter's newsletter product?
And here you see the guy who's just quit a CEO telling everybody not to use it.
So the company's just always been so frustrating and the product's been so frustrating.
Whether Elon is the guy to fix it like nothing he said so far suggests he has a clue how to fix it.
He's just sort of spiraled off kind of completely tangential irrelevant things like bots and freedom of speech, which we basically just suggest he hasn't really thought about like how hard content moderation really is.
The real thing is to have an org that makes product. It's not random product ideas.
It's to have like a fundamental conception of what it is you're trying to do.
And I did a podcast of mine about this and I thought, you know, comparisons with Steve Jobs are kind of desperately overused.
True. But look at what he did when he went back to Apple.
What he did was he first of all, he kind of he was there.
First of all, the operations and the product line was a total mess.
And secondly, they lost to the PC. And so he sorts out the product line and the operations and gets you go from a kind of chaotic mess of dozens of products is very clear, simple vision.
Which Mac should you buy and what is the Mac? What are the Macs for?
It was simpler. His vision was simpler. What is a Mac? I guess make a really clear proposition if you want a Mac, what it is and why.
So he makes it a functional organization, but it's still as opposed to a dysfunctional organization, but it's still lost to the PC.
So then it's a series of well, let's try and work out.
OK, is it video? So there's a period of video and clearly in some ways you could say the web saved Apple because suddenly there was a reason to buy a PC and it didn't matter.
There was no software, there wasn't much software for it.
But, you know, then it becomes the iPod and then that becomes the iPhone.
But there's two parts to this. One of them is like unfucked the company. Can I say that?
Of course you can. Yeah. Apple was a completely fucked up company and had been a completely fucked up company for a long time.
And Steve basically comes back and tries to fix the function of the company.
But that doesn't alter the fact that you've got 5% market share.
Exactly. He went then to the other stage. Then you have to do something else, which is, I think you said, you know, I'm going to fix it and I'm going to look for the next big thing.
You think Elon Musk can do it? Yeah.
So this is the thing, like you've got to like, is it enough just to get Twitter to be less dysfunctional and get it to the point that it ships products that work and have delight and joy to them, which they never really have?
And they're kind of integrated and joined up and make sense and aren't buggy and weird and don't make any sense.
But then the other side of it is like, OK, does that get you to the $15 billion from $5 billion to $50 billion?
I mean, it's really striking how small this company is.
I mean, it's got, what is it, 200 million DAOs?
They've got this weird metric, monetizable DAOs. What does that mean? Daily users, right?
Yeah. But monetizable daily active users. OK, anyway, so they've got 200 million DAOs and 38 million in the US.
It's like Snapchat three times the size in the US.
It's a tiny company. It is. It's interesting because it has a lot of attention.
We're giving attention, everyone gives a lot of attention. It's like the music industry.
Music industry globally, recorded music industry globally is now and it's been growing fast on the back of streaming.
Last year, I think it was a bit over $20 billion.
Yeah, that's a short industry. Very small. Yeah, it's like that's the change you find down the back of the sofa at Apple.
That's the cafeteria budget.
Yeah. That literally is if you've been to the Apple cafeteria, that is the cafeteria budget.
It's a good comparison because the influence is there because Twitter has a lot of influencers in terms of government people, journalists.
I was a journalist for 20 years. Journalists take a lot of content from there.
The content from Twitter goes to the news, goes to the cable news, goes to everywhere.
It has a big impact there in a sense. It does, yeah. It has that cultural impact.
It's interesting cultural differences of people in the UK and Europe pointing out here we've got this huge systemically important media company changing hands with no oversight.
In the UK or in Europe, if you were to buy a company with that kind of media voice, there would be government scrutiny.
There would be a sort of fit and proper test.
You're a fit and proper person to own this.
This is why Rupert Murdoch was unable to buy Sky, to take full control of Sky whenever it was 10 years ago.
But because it's the US, things don't work like that.
Which is kind of ironic given that the US is only 5% of global Internet users.
You get these kind of weird conversations where you have a conversation about content moderation and people say, well, this is what the First Amendment says about free speech, so that's the answer.
I don't actually care a bucket of warm spit what's in the American Constitution because I don't live in America and neither do 95% of people in the rest of the world.
That isn't an answer to what free speech is.
Free speech does not begin and end with one clause. Discussions and definition of what free speech is and how it works do not begin and end with one clause in a 1-200-year -old law.
There's a lot of other ways of thinking about what free speech is, which I'm afraid Elon is going to find out if he actually believes any of the stuff he said about this stuff, about which I'm kind of doubtful.
True. One of the interesting things is that Jack Dorsey backs this operation, has his own ideas.
He thinks Elon Musk can provide an open part of Twitter.
There's certainly an argument that says the board is dysfunctional and having to try and hit quarterly numbers is a problem.
Jack Dorsey said that Wall Street run the company for a period, so he didn't want that.
Well, you sold it. Exactly. It's public.
If you sell the company to people on the promise that you're going to give them financial metrics, then you give them the financial metrics.
He regrets that actually, Jack Dorsey.
The main regrets was putting Vine down. He regrets that, Vine going down.
And also putting Twitter in a public matter, but he has a particular philosophy there.
Yeah, he has a very individual view of what's gone wrong at Twitter, I think.
True. I think even Elon Musk doesn't go as much. He's been running the company for a long time.
At a certain point, you're the CEO. It's your problem.
Exactly. You're the CEO of the fucking company. There's some responsibility there.
True, true. Before we wrap things up, if you can, I would like to ask you about subscription models for consumer-based like streaming, Netflix, Apple, Tesla, Platinum, but also from the business-to-business side.
In terms of what do you think of subscription models, for example, now that Netflix is having some troubles, is having more competition, people cannot subscribe to everything.
In the business side, but also in the consumer side. This is a joke. What I want is for somebody to take all of these TV services and integrate them into one box that I could buy in a store with a cable that comes out of the back and I could just plug it into my television.
Like a bundle. You could call it TV by cable.
I mean, look, what's happened is that technology sort of changed the landscape of TV, but technology was unable to take control.
I always like to remind people that Apple launched the original Apple TV at the same event as the iPhone back in 2007.
One of those did a bit better than the others.
And Chromecast is the same kind of problem, which is basically what tech wanted to do was to turn TV into YouTube or Spotify.
So you would have one universal library with everything and one payment, and then they would control the whole thing and control the UI.
And then they would like divvy up the payments to the media companies out of the back door.
And that would probably be quite good for consumers.
But as a media company, as a TV company, why the fuck would you agree to that?
Why on earth would you let them do that?
Of course they didn't. And so Apple spent sort of 10 years trying to get TV companies to do this.
And TV companies were like, but why would we do that? I understand why you would, that would be good for Apple.
I understand why that would be good for Disney or Time Warner or CNN or anybody else.
So of course they wouldn't agree to that.
And so, of course, what actually happens and is sort of the economic.
And so the core of it is that software doesn't actually change the economics of making a TV show.
It just changes the distribution model. So what we now have is a model in which all the people that were previously B2B businesses going to market through cable now are now trying to become B2C companies going directly to market on these open platforms.
Of course, not all of them will succeed and there won't be room for all of them.
It's the same thing in D2C, just because you want a direct relationship with your customer doesn't mean they care about you or remember what brand you are at all.
Whatever brand you are, the consumer is the most important thing in your life, but you're not the most important thing in the consumer's life.
They don't remember which one you are, you're just the orange one, the blue one.
I don't know. They can switch and if it's monthly, it's easier to switch.
You can see one month. Yes, this is the other thing.
So the thing about, you could argue that this whole sort of phrase content is king is almost kind of an anomaly because there's this period where you actually have to buy the content.
So you're buying the origin of the phrases from VHS versus Betamax.
And the content is on one or the other. You kind of have to make a choice. And the same thing then with music, which is DRM.
So you bought 50, 100, 500, $5,000 of music on whichever device, probably an iPod.
You can't switch to another device because then you'll lose all the music because it's a different platform.
But once you've gone to streaming, you don't lose anything if you switch.
I mean, if you switch from Apple Music to Spotify and you lose a playlist, basically you don't really lose it.
You don't lose 10 grand of music. It's just a bit of inconvenience. And so when we went to streaming, the kind of content as strategic lever for tech, the value of content as a strategic lever for tech basically disappears.
And so this is why Apple is now got such a marginal role.
And frankly, Google sort of marginal role within all of this.
Of course, where Google is hugely important is in thinking about converting TV advertising, which is not at all the same as converting TV.
So the advertising is like half the advertising industry and it's the half that hasn't moved yet.
So that's the next 10 years of Google. But now within that, of course, do you go ad funded or do you go to subscription?
And then there's a whole bunch of maths.
This was kind of fascinating to look at Spotify, which has got a tiny advertising business.
The advertising business at Spotify is basically just there to annoy you, to get you to take out a subscription.
I mean, it's a loss leader to make the free trial annoying. The ad supported Spotify is actually a free trial.
And the ads are there to make it annoying so you pay.
Exactly. It's the entropy. Yeah. On the other side is, you know, advertising is, you know, the whole of advertising is kind of thrown up in the air.
Both on the Internet side with cookies and privacy and everything else and, you know, the disconnection of TV.
So all of that. And then we have this kind of great wave of creation of new Internet brands.
So it's sort of in some ways tremendously exciting, but sort of terrifying time to be in advertising at the moment.
Where does all this money go and how does it move? And how much of it goes from advertising to rent?
And how much money goes from rent to advertising if you don't need a store?
And how much do you decide maybe we should have stores instead of spending our money on Instagram?
So all of this money sort of moves around in all sorts of different ways.
And a subscription TV is kind of just one part of that kind of vast continuum of what the hell's going on.
I think clearly, you know, the sort of specific Netflix questions, which is.
You know, there's one certainly one view of Netflix would be there's just too much stuff that's mediocre.
And you open the app and you think, why am I paying for this?
And you could also almost, I think, argue that there's a sort of diminishing return.
So on the one, first of all, I think it's pretty clear there is not a network effect where it takes all of that.
It's not like Uber or Instagram, where the more people use it, the better it gets.
It's a scale effect, which is the more people use it, the more money they have to buy more TV shows.
But that's not quite the same thing.
And after a certain point, more TV shows doesn't make it better. In fact, more TV shows might actually make it worse.
Because now it's just the level isn't top right.
Well, there's sort of two things within this. There's an obvious point, which is you can't find the diamonds.
But there's a slightly less obvious point occurs to me, which is maybe if you've got 50 great shows, then there's none of them that are must watch.
There is no show that everyone is watching. So you must watch because there's too many of them.
So there can't be a show that everybody's watching.
And so it's just a hypothesis, sort of a diminishing return, or even maybe a negative return on increasing programming investment over a certain point.
I mean, Netflix spent sort of 15 to 20 billion dollars on content, depending on how you count it, which is more than anybody on us except Disney.
It's also more than all the European broadcasters and top five markets combined.
It's more than every broadcaster in the UK, France, Germany, Spain, and Italy combined.
It's pretty impressive. But there's a sort of a deeper point here, which is like, well, you know, how many people are going to pay how much for how many of these things as you come out of the pandemic?
And the deeper point within all of that is sort of, well, these are TV industry conversations.
They're not tech conversations.
You know, the tech is a commodity. It has to work. You know, streaming has to be good.
The app has to be good. But like if all Netflix had was I Love Lisey and Friends and Treasure, no one would be subscribing.
It's all about the TV.
True. Yeah. And the other is now Disney has Marvel has big names. Yeah. And if the app isn't quite as good in the streaming, like who gives a shit?
Actually, that's not what matters. What matters is it has to be, it's a condition of entry.
The app has to be good. If the quality is nearby. Yeah, the app has to be good.
But that's not what, that's not your differentiation. The differentiation is what's the TV and the TV is not a consequence of software.
The TV is a consequence of how much money you have and what relationships you built.
And your talent in creating stuff.
But fundamentally, it's, you know, it's a Los Angeles question.
It's not a San Francisco question. So the thing that I've written is, you know, technology, you know, that, you know, the same thing in DTC.
You know, is that a bad company or tech company?
Is Woobie Parker a tech company or is it a bad company?
Is it a boss's company? What are the questions that matter there as a retailer?
Absolutely. I have this question for you. It's like advices for those in tech that do product, product managers, or even CEOs.
I give you this question. Why? Because even at Web Summit, we were talking about Web Summit.
I saw some guys were talking about how tech CEOs like Elon Musk are seen as the new era philosophers.
They are philosophers in a sense, even for people.
Some people look for those guys. In the 20s and 30s, there's this thing called Fordism.
There's a whole cult of Henry Ford.
So this stuff always kind of comes in waves. And, you know, there are moments where, what's the Bismarck line?
You know, the great man is somebody who hears God's footsteps and grabs onto his coat as he walks past.
And so, you know, what Elon did was sort of realize that basically a lithium-ion batteries driven by smartphones would make it possible to make a good electric car.
As opposed to a lead-acid battery, which produces a shit electric car.
And that if you were to make it, you could make a good electric car.
And if you were to make a car that, I think, John Lewis Gasset, who was at Apple a long time ago, said a Tesla is a car for meat eaters.
So it's not a car for hippies. With a lead-acid that drives, does naught to 60 in two minutes.
You know, you could actually build a car that's electric and a car that's actually good.
So he sees that and he grabs onto it and he makes a hat.
And the same thing with SpaceX. You know, he sees that it would now be possible to make reusable rockets and the economics of that might work and he makes that happen.
Steve Jobs saw that it would now be possible to make a piece of consumer PC, that it would now be possible to make a smartphone that was good.
And he made it happen. And now, of course, there's hundreds of people around that that actually turn it from the force of will into the reality.
But you kind of at certain moments, like you need that force of will to make that happen.
The other German philosopher quote, of course, is Marx, who after spending, you know, 10 ,000 pages talking about vast impersonal forces driving history forward, then says history is not a person.
Somebody still actually has to do the revolution. True. That's a good one.
In terms of advices, if you have to give advices to product managers, those who are building products, do you have any advice?
I've never built, I've never run a bath.
I have no advice to build products and how to build. But you will analyze those products when they're out there, right?
Well, yeah, maybe. I mean, what's the Groucho Marx line?
These are my principles. And if you don't like them, I've got others.
That's a good one. Just to wrap things up, we were talking before about subscriptions.
There's a lot of talk about bundles these days. Big companies are bundling things together.
Where do you see bundling going? As I said earlier, and I'm going to wrap up on this point.
But as I said earlier, there are these sort of various different kind of taxonomies of cycle and tech.
And there's a client-server cycle and a open-closed cycle and a bundling-unbundling cycle.
And depending on where you are in the process, there's always a point at which it's more efficient to bundle and a point at which it's more efficient to unbundle.
As we were just saying, unbundling cable into all of these different services.
At the moment, that's where there will be a point where it gets bundled up again.
You could look at many enterprise software companies that are basically unbundling Excel or bundling Salesforce or bundling SAP.
But then, of course, many of them are bundles in their own right.
They bundle up things on some different axis. They bundle up image recognition with messaging, with workflow, for the sake of argument.
They bundle up on a different axis.
And so that cycle just kind of always goes round and round.
But do you think that cycle will expand, for example, doing like iPhone bundling type of thing, subscription type of thing?
Even Tesla, you can make a subscription about your car, in a sense, bundling.
As I said, these are just endless cycles.
Every company, there's always the point at which somebody says, that product is huge and bloated, and I'm going to peel off this one use case.
That big general purpose product is huge and bloated, and I'm going to peel off this one specific use case and nail that.
And then I'm going to add this, and I'm going to add that, and I'm going to add this, and add that.
And 15 years later, you've become what you killed.
Oracle hated IBM and became IBM. Salesforce hated Oracle, and the Salesforce has become Oracle.
Salesforce in 2000, it's just a URL. It's incredibly simple and easy to deploy.
Salesforce in 2022, well, we've got a three-year deployment roadmap, and we've got 150 Oracle-certified engineers building our deployment.
Well, yeah, you turned into Oracle. Although there's evolution, right?
There's evolution. It's just an endless cycle. Microsoft has become IBM now.
Google has become Microsoft. It's interesting. It's really a cycle, history cycle, in a sense, like you were saying.
For the future, what are the trends you're seeing more promise?
I don't know if you're a VR type of metaverse. I don't want to get into another kind of 20 -minute conversation about VR.
I think we kind of talked about this.
I mean, we are in this point where there was that one megatrend of mobile that sort of mostly played out, and the cloud megatrend and machine learning megatrend were kind of halfway through.
And we are looking to see, is there a VR and AR megatrend or not?
Maybe. We don't know. Is there a Web 3.0 crypto megatrend?
Yes, but what? I mean, I generally compare crypto and Web 3 .0 with open source in that this is a new architecture, a new way of building software.
But, okay, well, what role will that play? How universal does that become? We don't know.
And meanwhile, we just carry on deploying all the stuff that we've had for the last 30 years, except that we haven't finished yet.
Just to wrap things up, are you excited about one of those techs that are coming?
I think I'm always interested in new questions.
I think in technology, the moment you understand something, it's probably the moment that you should be paying attention to something else.
Interesting. Thank you so much, Benny. Hope you liked our conversation.
Thanks a lot. Thanks. Bye -bye.