Voilà, a volume on quite impressive volumes, do we still need developers? And when I say developers, I could actually be talking about all the technology professions.

What I would like to note, however, is that it’s not the first time we’ve had a technological revolution, or at least a technological evolution in the field of IT. Fortran in 1957, the very promise of the thing was to say, you’re going to be able to create functions that will give you a given result and you won’t have to code them in assembler. Assembler is a bit annoying. Generally, you had to uh unstack information in a register, apply a set of operations uh and then put them back in a register and then make jumps too. It’s a notion that there’s an iteration to be done on that operation. In short, it was a language that was quite complex, let’s say, to put in place. And the promise of Fortran is to say, we’re going to be able to lower the entry barrier and we’re going to be able to allow scientists, in particular, to carry out their own research work with a higher-level language. That’s simply it, when I say low-level, high-level, it has nothing to do with the level needed to code. In any case, has the developer profession disappeared? No. But, it has indeed changed.

There are other changes like that. SQL, likewise, is also a change in abstraction. which allowed us to retrieve data from a database. And here again, the stated objective is to say, we need to open up our profession more to other people who are likely to do it.

MDA, does anyone remember MDA in the room? Yeah, that made an impression on some people. We were two steps away from an industrial catastrophe, which was that we could have coded everything in UML today. I’m very happy we’re not there. Oh no, but the promise in 2001 was to say, we’re going to start from the UML diagrams that are set by the objective, which is to say, the functional teams will be able to create all applications directly from their UML diagram. It worked. but there were still quite a few initiatives that were done at the time and which were, all things considered, quite interesting. The only small drawback is that it didn’t generate, it didn’t allow managing the post-life of the application, it only allowed managing the start, i.e., the initial writing. IDEs, I’m going to go very quickly, but anyway, people who were in the middle and people who didn’t use them, it was an additional level of abstraction. I mentioned outsourcing because I find it interesting to say that, similarly, by saying it’s the end of developers, when in reality, it was about having developers in another country. I find this thought quite amusing because it reminds me a lot of I don’t know if you’ve seen Musk’s androids, the androids that will be commercialized starting next year, I think. Or at least, that’s how it was planned, and you never know with Elon Musk. Anyway, Musk’s androids, in fact, they are not at all autonomous. In fact, they will rely on people who will pilot them and who will be, a priori, I believe, in the Philippines. so there you go, I hope you’ll be very happy to have lots of cameras at home watching your dishwasher, your laundry, etc. We always have that little feeling of saying, ‘Ah, we made a profession disappear.’ Well, in this case, outsourcing didn’t really work either. You’ll notice that we still have people working here in France.

The code, the code, I won’t insult you by reminding you what it is. Again, is this the end of developers? No, it wasn’t. Why? Because, ultimately, our code, well, our job, is not solely to write code. A company is not limited to the code it produced. And writing code also requires knowing, in quotation marks, how to fall in love with a problem, making compromises, making design choices. Design choices that are by no means neutral, by the way. These are things I’ll talk about later at another time, in another place. But the fact remains that we do indeed need, and AI cannot do that. AI cannot fall in love with a problem. AI cannot make its choices and its compromises because that requires having a fine knowledge of the company, of its economic context, sometimes even of its political context, in short, of the application that depends on it. In short, there is still a need for supervision. So, is this the end, is this the end of no? But it’s true that I can’t tell you today if the name of the profession will still be developer in 5 or 6 years. It is probable that our profession will change. When I say our profession, again, I’m including all the people who are in the software production chain. I think all of our professions can potentially change their name. I don’t know what to do yet, anyway, that’s what we’re going to talk about. I want to talk to you about the cultural impact that a technological change can have. It’s not the first time we’ve had this. And so for that, we’re going to go back a few years in the past, we’re going to talk about the cloud. This famous word that really crystallized the debates for a few years. when I started Malt, it was in 2012. We created a company that we call cloud native. So we’re talking about a label, we’re not talking about the substance, we’re saying we created a cloud-native company, simply, we took the best practices of the time and above all we tried to do something in opposition to what was done before, that is to say the pre-cloud period, let’s say. And the pre-cloud period, I knew it well, there were things that could sometimes be a little irritating. For example, I don’t know, in a company, I once asked for a CI server, so a new machine to be able to do tests. It took maybe 4 months for me to get that machine, between administrative requests, extremely painful bureaucracy, opening flows right and left, etcetera, etcetera. to say that obviously, these are the kinds of things we wanted to avoid a bit with the cloud, we suddenly arrived at something where with one click in an interface, I could have my machine. But that’s only the technological part. That is to say, in fact, when we built Malt at the time, we were able to retrieve our HR service, our payroll service, our accounting service, our messaging service, uh all these building blocks, we assembled them from the cloud. So we ended up with a kind of giant Lego. which uh well, which represented a new form of company, a dematerialized company, serverless, something quite different. Maybe someone who I know that often the cloud is considered a bit like a marketing gimmick in the sense that it’s a managed service whose name is not mentioned. I consider that to be a mistake. There is a real cultural change that has occurred in companies uh regarding this technological change, and what has allowed the generation of startups that we have found, among others, in the last 15 years. The speed that was offered to these companies through this technological change is important. And that’s where we realize that it’s not a technological change, it’s really a change in approach. Moreover, in a somewhat concurrent way, you will notice that methodological and cultural approaches have emerged in companies. I don’t know if they were born with it, but they were reinforced. Think about the DevOps change. Think about continuous delivery, again, it was continuous delivery before, but there was a reinforcement with a whole set of technologies that were linked to the cloud. And one more thing, the end of 6-month release cycles. Finally, I was in a company, generally, there was a production release every 6 months, or even sometimes more, in a banking context. That’s something we’ve seen disappear a bit, and it has cultural implications. So, and just to insist on one thing too, when I talk about cultural implications, there are companies today, even 15 years later, that are still not what I would call Cloud Native. For the simple reason that a technology is adopted, while a culture is implemented, and sometimes the grafting doesn’t take. That’s why we’re not there yet. So would it be completely absurd to say that AI uh will bring us a cultural change and that this change, that this technological change will be accompanied by a cultural change? I think that would not necessarily be absurd to think, at least given what we’ve seen in the past. And I would even draw a parallel between the changes that will be linked to AI and what we saw with the web or mobile, as the implications affect not only companies but also the general public. So I’m going to call it the non-enterprise, native AI.

One of the first visible elements that we can see today when we talk about AI is a very, very simple thing. The quantity of code, the volume of code we produce, is no longer a natural protection, no longer a competitive advantage that we can have externally. an English word, a moat towards the outside. And I want to make you think for 2 seconds, in your opinion, How many people today work on LinkedIn? So, who is making the LinkedIn application, how many people today are making this application? Do you have any figures?

taken, there’s the church. One of the first visible elements we can see today when we talk about AI is uh a very, very simple thing. the amount of code. The volume of code we produce is no longer a natural protection, no longer a competitive advantage that we can have externally. A kind of English, a moat, there you go. Which protects from the outside.

And I want to make you think for 2 seconds. In your opinion, how many people work on LinkedIn today? So, by building the LinkedIn application, how many people today make this application? Do you have any numbers? 2006, OK. Why 6, people? Ah 2006, I thought that was a bit small though.

Something else?

3 million. OK.

the real answer I was given is 19,000 people. Out of these 19,000 people, we have 15 billion in revenue, if I’m not mistaken. So that. There are between 8 and 9,000 people working in tech out of these 19,000 people. So when I say tech, it’s the product in general, all professions combined.

And yet, we can reproduce, so I’ll explain why I added the screenshots. we can reproduce the application. There’s a person who had fun doing it, the application we see here is Ponos. So it’s Palo Pernaud who launched it not long ago. and who had fun reproducing LinkedIn, but he’s all alone and he notably acquired Colette. Here, I put another application, it’s a Meetup equivalent, I talked about it with someone, I don’t know if Drago is there. So uh Drago also had fun rebuilding something that’s a kind of Meetup equivalent, if I’m not mistaken. So I added the screenshot, as I told you. to give strength to all these applications, not in relation to AI, that’s not the subject. but in relation to the fact that these are European alternatives to American platforms, and today, I think we have a real issue from that perspective. It lacks a bit of network effects, it lacks a bit of participation, but all this to say it wasn’t. Will these applications immediately supplant LinkedIn or or. It’s obvious that it’s not that simple. So the quantity of code is no longer an asset, but there are still limitations, meaning that uh these companies, they still benefit from other things. Now, that doesn’t mean that

I worked for a company in the finance sector a few years ago, where globally the amount of code is a company that was built, it’s a bit strange, 60s ex-software that was used at McDonald Douglas, the accounting shell at McDonald Douglas and that has evolved so much that today it has become a software that values I don’t remember the percentage but more than half of the financial flows that work in Europe. Uh so it was a company that had the monopoly at that time on the valuation of financial funds.

And there were millions and thousands and thousands of codes. It was impossible for a newcomer to arrive and say to themselves, I’m going to be able to immediately dislodge them. Uh in any case, never by surprise. I think that this situation has just changed because indeed the volume of code today no longer matters.

We’re going to talk about a small thing. I think this person I don’t know if everyone recognized him, it’s Darwin. And you’ll quickly see the connection with what I’m going to tell you. Many analysts have started predicting a huge crash on all the SaaS, so SaaS, these are the online services that have benefited quite a lot from the period of the last decade. because precisely this cloud culture, this culture of consuming uh software remotely uh every time we grow, you just need to add a little money to the credit card, that’s how they grew but now, these services are in danger. It seems that there is, so we’re talking about SaaS apocalypse, it’s good to bring some dramatic words.

285 billion in market capitalization have already evaporated. to the native. It makes me laugh because not long ago, I heard that Anthropic was releasing a new module to potentially manage code conversions uh from Cobol uh to more modern technologies, and uh following that, IBM’s market capitalization drastically fell. a bit like the company I told you I worked for just before. I think there’s also a rent that has formed, so today we talk about that, we talk about is AI today, rather than eliminating jobs, not eliminating rents? I think indeed, there are quite a few software rents that have been built on volume. And there will be some questioning about that.

To measure that, there’s a stock market index today that exists, called the Morgan Stanley SAS Index, which as you can see, is starting to decouple a bit from the Nasdaq. So, would this be the first time, well it’s not the first time, sorry, that we see a stock market index that doesn’t reflect the real valuation of companies. we agree on that. However, it would be interesting to ask ourselves what could justify that.

And I’m going to come back to something, it’s uh I think you’ve often heard the ‘buy or rent’ dilemma. in 2003 I was very new in my career and I was doing training to become an architect.

And I was told, among other problems you’ll have to learn to solve, besides naming and cache management, as everyone knows, these are two more important problems in IT. Uh but there will also be that, the ‘buy or rent’. you’ll have to learn to make a distinction between what needs to be built and what, on the contrary, needs to be bought off the shelf for economic reasons, also for focus reasons, opportunity cost in fact.

But a priori, from now on, we have a new constituent which is what we build, what we buy or what we rent. It’s true that, put differently, we have another dilemma today. Rather than buying a SaaS for 500,000 euros per month, wouldn’t I have someone from my team work for a day to try to reproduce only the functionalities I need but tailored for me? In some cases, the economic equation will be interesting. Not in all, and in some cases, it will arise.

this is one of the reasons that explains the risk that can bear on all companies today that offer online services. Now, there’s a second subject. I told you earlier, LinkedIn 18,000 people, 10,000 devs. Ponos, one person. Most SaaS today bill per usage.

So imagine the total valuation, well the resource park that is valued in companies, between one person and 10,000. Of course, there’s also that, most SaaS today that bill per seat, well, they’re a bit bothered if, from the moment companies become more lean, they start consuming fewer resources. Well, don’t worry about that.

Uh I remember, a long time ago, uh Oracle used to bill uh I don’t remember uh by the number of connections on databases, and then uh we invented connection pools, and Oracle, they knew how to adapt in terms of pricing changes. So on that, despite everything. that’s part of the things, so we need to ask the question of the alternative. Now, can we really consider that uh SaaS will disappear? I told you earlier, LinkedIn, will Ponos replace it? The good thing is that it’s not that simple.

Because there are still advantages that can be important for companies. And I invite you, by looking at the advantages that are there, to reflect on what you have in your respective companies. What are the things that protect you today from a new entrant, from native AI companies that potentially will go faster and we’ll see why later. First thing, exclusive data. For example, if I rebuild Netflix, do I have access to their catalog? No. So indeed, I don’t have the data today, I can’t obviously rebuild a Netflix. Do I benefit from a network effect? I don’t know if you’ve ever tried to leave WhatsApp. It’s complicated alone, you know.

Inevitably, we stay there. The network effect is that, a software that has value because there are people who use it. Certifications, the fact that you have, for example, secret defense accreditations in the military. The fact that you have PCIDSS in the banking sector. All these things, these are not things that are replicable in a software domain, there are real organizational issues that are behind all this. So that still protects a company.

The ecosystem.

The fact of saying that, for example, a software is quite strongly implanted in an entire software layer, an information system in the company. Think of Jira, think of Salesforce, think of Okta, all these software, of course, you know it’s very difficult to dislodge them. Even something like Jira, we would say OK, it’s a ticket manager, I could handle it, it’s easy. Uh in fact, often, it’s connected with an HR thing or time management, for example when you do Scrum or uh you have examples that make it not so simple in reality. Captive data.

Uh that’s a bit different from exclusive data, it’s the fact of saying that you have data, it’s yours, but uh you have an application that will modify it and then will make it captive, well the transformations of this data will be made captive. Think of Palantir.

Uh which is, you know, the absolute evil right now, I hope no one among you uses Palantir.

an American software that does large-scale surveillance. But anyway, that software, it has captive data, it’s not easy to dislodge Palantir.

SLAs, everything that is service guarantee. Many people, myself included, are ready to pay for a service guarantee on certain critical systems. Uh typically on banking. It’s not just the fact that I don’t have the PCI DSS application that makes me not want to reproduce Stripe. Okay. If I want to use payment software, I prefer to have people who guarantee me that it works. Uh it will be the same under cloud providers, it will be the same in many other areas. Globally, I want there to be people who know who sign in their blood and with penalties uh that it works or it doesn’t work. The physical network. Think of Amazon, whether it’s servers on the AWS part, whether it’s logistics, warehouses, the commercial part. And distribution, finally. Distribution is uh your ability to distribute your product. Let’s imagine that, for example, I rebuild Teams or Slack, OK, but I don’t have the commercial force of either Salesforce or Microsoft to then distribute it in companies. I didn’t put a specific thing which is rather protected sectors, that’s basically when you have a monopoly that guarantees it, obviously I could have added that.

Anyway. So, is it the end of SaaS uh and is it generally the end of all current businesses? No, as you’ve seen, there are still quite a few other protections that allow you to shelter yourself from that. Now, that still raises questions. Uh that should make us all ask questions about this kind of business. What do we sell, what protects us? If it’s just the amount of code, then there’s a real issue. I don’t know if I give you a very simple example, if you use Launch Darkly, who knows? Launch Darkly, Unleash, it’s a software that allows you to do feature flagging. Uh a feature flag, it’s an ‘il point elle’, it allows to condition, well I’m simplifying a lot but it allows to globally condition the activation of a functionality at your place. It’s not a very complex software, but normally we don’t react like that, normally we tell you OK, it’s not a core for me. not a core for my company, so I won’t recode it. Now, if with uh Vibe coding, I’m able to reproduce the application in one day with only my own needs, what we can ask ourselves. So you all have a question to ask yourselves about the competition as it will appear in your companies.

Ah crap, so that makes us do the whole thing again.

So, we’re going to come back to uh what I call a bit the AI Iceberg. OK, that’s cool, we just saw something. It’s that the first the first subject that seems quite obvious is the cost of code. To defend because in fact there are other applications behind all this. The first one is uh I don’t know if I should have called it that, but it’s really the fact that we’re going to have to re-think our software factories. We’re going to rethink the way we do software development. When I talk about software development, in reality, I imply the entire chain, from iteration to production and so on and so forth.

these words, I don’t even know if they will still be valid in a year. There have been a lot of changes over the last 20 years. If I generalize, I’d say, we went from an evolved autocomplete around 2021 to people who were doing a bit of tedious copy-pasting in a browser in 2021, 2022, 2023. to agents to whom we can completely delegate the entirety of our work since 2023. The word ‘Vibe coding’, the word ‘Vibe coding’ only dates back to February 2023. So there was a real pivot in 2023. That means that behind it, we are facing teams that must entirely ask themselves how to reindustrialize all this, how to regain control, how to make it reliable. How do we test that it actually produces what we want it to produce? Uh and so on and so forth. So that’s there are quite a few people working on it, so I know some in the room are working on it since we already discussed it. That raises another question behind it.

No, we’re going to tackle, we’re going to look a little more at the individual impacts, it’s loss of bearings. Many people are asking themselves, OK, but what’s my job then? Because I don’t know if you represent uh I don’t know if I imagine that you represent yourselves pretty well. When one has built most of their career on building expertise in a domain and suddenly uh there is uh a terminal that arrives and reproduces what you do, uh that stings a bit. If suddenly our job, here I’m talking about people in in, well if they develop software, if suddenly the job becomes only reviewing PRs. Ah is that really what I signed up for? If tomorrow 90% of my day is meetings and reading tickets, as the other would say, I didn’t come here to suffer.

There you go.

So from the moment where indeed uh there is there is there is I I completely understand this feeling, but it’s true that I want to bring another perspective.

In reality, for my part, I find it fun.

So, what I mean by that, I’ve been building applications for a long time and in fact, I’m quite happy today to be able to quickly get to the result. Until now, I’ve always had an “idea tax”, an execution tax, meaning if I have an idea, OK, but it might take me time. Uh if I have several ideas, I’ll have to make a strict choice because in fact, I won’t be able to look at everything. Now, I no longer have this “idea tax”, meaning I have the idea, the test.

And so for me, in any case, I find it quite nice because not only can I think much more about the ideas I have, but I can put them into practice much faster.

Now, I’m telling you this, obviously I still have an advantage, which is that I spent 25 years making applications, knowing how to build them, so it’s easier for me to say there were things that weren’t super nice. You can imagine that I was a bit fed up, I won’t hide it, with making forms, well, and consistency checks in the data. So I’m very happy to get rid of those things.

But I can understand that it stings a bit more when you have, for example, between 5 and 8 years of experience. because it’s the moment of fermentation, of intellectual maturation, it’s the moment when we have acquired technological skills. And suddenly, they come and take away a bit from us by saying it’s no longer useful. it’s not perfect, it’s not useless, but that’s a bit of the message that’s being sent, obviously, it’s a bit difficult. So in companies, there is a real issue today which is cultural. How do I support all the people, I know quite a few seniors who are in this situation. And in companies, there are also people there to go and see them. I don’t know if you’ve already started to see this approach in different companies. It won’t be something that will happen in so-called native AI companies, because these companies were born with it. The starting condition is that. So, they will ask themselves adaptation questions. Whereas conversely, potentially, if you are in a larger company, you will have to do it. You will have to set up cells when I say cells, we’re not talking about psychological cells. Uh there you go, but you’ll have to set up accompaniment mechanisms so that people can re-understand the job a bit. On the other hand, that also means that we will have to reposition all the professions. We will have to re-examine the question of everyone’s responsibility in the software development flow. We’ll come back to that later because that’s not a real individual impact. it’s already a bit about having this approach in the different companies. It won’t be something that will happen in so-called native companies. Because these companies were born with that. The starting condition is that. So we’re not going to ask ourselves questions about adaptation. Whereas, conversely, potentially, if you’re in a larger company, you’ll have to do it, you’ll have to set up cells. When I say ‘cells’, I’m not talking about psychological cells, but you’re going to have to put in place support mechanisms. so that people can perform reasonably well in their jobs. However, it also means that we will have to reposition all professions. We will have to re-evaluate the responsibility of each individual in the software development flow. We’ll get to that later, because that’s not a real individual impact.

Another one. Oh, that’s great. Another topic I wanted to talk about, and here I think we’ll all agree that it’s pretty good. The end of ego programming. So ego programming, for those unfamiliar with the term, is the somewhat unhealthy attachment one can sometimes have to their own code. which is perfectly normal, if you’ve worked for 2-3 weeks developing something, you’ve put 13 hours of engineering into that famous code. Well, all of a sudden, you’ve kind of developed, a famous Stockholm syndrome, actually. which means that globally, you’re going to negotiate with all the people, to tell you that no, I don’t want to change the date you told me because it actually already cost me 2 weeks. today, I don’t want to continue. I think you’ve all experienced that. sometimes we’ve done it, I’ve done it myself a bit in bad faith, a little bit. Is your gain really sufficient for me to willingly make the modifications compared to the effort required? AI will be an ego neutralizer, meaning I am capable today of deleting 1000 lines of code at once. because I don’t care, I didn’t. And that’s still going to change a lot of things. Today, I am much more capable of telling myself, no, the result isn’t perfect. Before, I would have said, it’s not perfect but it’s okay. Now, I have it’s not perfect. No problem. I’ll get rid of it, I continue. And I think that’s going to change software development habits.

Another thing that’s closely related to this is, I could have put a second item, but for me it’s the same thing. gatekeeping, gatekeeping, if you’ve seen it, is it a kind of very toxic component of ego programming? It’s the fact that some people start putting up barriers around certain parts of software, in this case, and who, do. Ah, who throw a little black magic in the air to say, well actually, no, only I can understand, not you. that’s going to be complicated to justify now with AI. because first, it’s not the people who will write the code, and then also because AI can react at some point. So all in all, I’m pretty happy, to be honest, that it’s going away.

However, of course, I added more. I made three three slides each time. However, there’s one thing that could potentially affect you, and that is, well. I saw an article not long ago that talked about AI and vampires. I find the word quite interesting, so we’re going to focus a bit on that.

There was a study by the Harvard Business Review that came out not long ago, which explains that, in fact, when we have tools that allow us to go faster, we might think we’ve saved time. And the reality is that work tends to intensify. This intensification will come with several consequences.

Fatigue, burnout, error, something that is really documented in my research. There are several reasons proposed for this. There’s one that interests me because it has implications behind it. The reason put forward for this, or one of the reasons, is to say that before, there was a set of tasks that was the production, I told you, things that tended to annoy me after a while in my career. but despite everything, these tasks, which were repetitive, repetitive, well, they gave the brain an opportunity to take a break. It’s a cognitive break. It’s the mindset of saying, ‘Hey, I’m going to start thinking about something else.’ because mechanically, there are things I’ve somewhat automated in my head. So I have them. And then my brain thinks about something else. And then it diverged. That’s also what we call incidental thinking. Incidental thinking is what we can have in the shower, in the toilet, to each their own, but it’s what allows us, globally, to suddenly help ourselves, to consolidate the information we’ve found. and after a while to have that little Eureka moment.

No, actually, AI removes those moments from us. So we end up with an activity throughout the day which is constantly making critical decisions. The only thing AI asks us is, I didn’t understand this thing. give me uh your solution for this thing, give me your choice for this thing. And we’re not wired to make 8 hours of critical decisions non-stop. If that’s all we do all day, we become a bit… So that imposes something on us.

It forces us to rethink our time management. If we saved time. Okay, but what’s the point?

So at Light, it’s very simple, there are two of us. So, in 3 months, we created an application that, would reach the level of 80 for example, 80 to 150. And yet, we took the gamble of not working full-time. in truth, we didn’t make a bet. In truth, the thing is that I’ve been working remotely for 15 years and I also know how we need to manage our mental and physical state regarding that. I’m not saying I’m telling you nothing about remote work, that’s not at all what I’m saying, but I’ve learned to manage the ports. And it’s exactly this skill that helps me work clearly today. I don’t do 100%. I code part of the day, and then the other part of the day, I do something else. I do woodworking. So there, I’ve created several pieces of furniture recently at home. I’ve improved on my dowels, I imagine not everyone is into that kind of thing.

Obviously, I know that’s not reproducible for everyone. Obviously, I know that in my company there are two of us and it’s very simple. You can imagine that starting a situation, a woodworking session in an open space can raise many questions.

But there’s a real question, indeed, a reflection to have on what do I do with the time saved? No, I saw in that article and in another one I read, it says in fact, you shouldn’t overproduce, the solution shouldn’t systematically be overproduction. Which is likely to be the automatic reflex we’ll have in a lot of companies. So I saw an article not long ago on GitHub, or rather from the company GitHub. who said that basically, they had changed their management of who, in fact, they devoted much more time to collaboration and ideation, but they had removed a large part of the production side. Which, in any case, is quite logical because production, in any case, will have a limit. We often talk about the rebound effect, but for the rebound effect, there needs to be a use case in front of it.

There needs to be a demand on the other side. Not long ago, I was working, well, I was discussing with a friend who works in video games, he told me that the number of video games has increased tenfold in 10 years. But the number of hours spent by players has not fundamentally evolved in those 10 years. 50% of games on Steam have less than 10 reviews. So consider that obviously between friends and family, sift through it, I think 50% of these games actually don’t even have players. If there’s no demand, there’s no point in overproducing. In any case, you have two problems to manage: first, the fatigue of the people you see. But second, are your users ready to endure 10, 20, 30 new features per day? probably not. Is your marketing department ready to endure daily notifications about pricing offers, about things included in the product, about training to be done with the support service, and so on and so forth? In short, we’re facing something where, in fact, we don’t have the capacity anyway to make our company overproduce to that extent. Some people will tell me, yes, but in fact, I already have a whole sick legacy to manage. That’s true. That didn’t last for years.

So yes, we’re facing a subject a bit like this one, where we’ll have to ask ourselves that question.

Now, let’s go down a bit to the team’s impact. I like this one.

Are we going to ask ourselves a bit about why we do ‘grand’? Or rather, why do some people do ‘grand’? I never had to advise to do.

But it’s important to understand that in many contexts, I think we’re all aware of this, as code takes a long time to produce, we’ve had a lot of solutions for that. The first thing we told ourselves was to put more people to produce it. Once we added more people, we also understood that it was expensive, but despite everything, for it to be efficient, we said, we’re also going to add other people because people need to learn to discuss together, we added managers. And then, in a team of 100 people, we said, ‘Okay, we’re going to have a team of 200, we’ll have to segment, we have tracks, we have squads.’ We naturally started wanting to align all these people. We have managers, managers. And then, we started having people who were only available for certain teams, a copywriter for example, you don’t necessarily have one in your package when you hire no one, so it’s shared, and so on and so forth.

We also started to put in a unit of measurement, the ticket. The famous ticket.

which is theoretically a kind of unit of flow measurement, on which we tried, for better or worse, to put units on it, we call that complexity. which is always hypocritical in a lot of companies because in the end we say 3 points of complexity is 3 days. Anyway, we tried to do that and then we tried to calculate the velocity of the teams. All this to arrive at what? A dependency diagram, we told ourselves, well, that’s how team X, it will perhaps need the copywriter at that moment, and then it will have infinite things, etc. Well, that’s all that starts to work.

Imagine tomorrow that certain companies, the native ones, will start with a fundamentally different structure, 10, 100, 300 people, obviously, they won’t have to deal with that tax. They won’t go faster just because they use AI, because in fact everyone will use AI at some point. They will go faster because they will not have this coordination tax. In fact, we’re talking about something here, we’re talking about a return to organizational sobriety.

And regarding this organizational sobriety, there are two things I wanted to talk about.

I experienced the growth of a company from 0 to 700, so when I talk about organizational sobriety and a return to that. Anyway, I’ll mention two or three anecdotes, but.

When we started to grow, this is Brox’s box that globally models the number of relationships we’ll have in the company and the number of possible connections between individuals as we add individuals to the group. Every time you add someone, it explodes the number of connections.

And it’s true that, at least from 0 to 100, I really felt that. 0, 10, 20, and 50 are roughly the four milestones I’d like to give when starting. 50, it starts to hurt. 50, it starts to hurt. Because not only is there indeed this explosion in the number of connections, which means we can no longer all talk to each other when there are more than 50 people. still very well, but. Ah, it’s complicated, you have to start putting a lot of things in place. And then, we also have specialists starting to arrive where before they were generalists.

The second limit is rather 100. For me, it’s marked 150 here. I would have personally said 100, it’s at 100 that I started to find it a bit more difficult. Because we don’t have the capacity. I can’t say, I’m not a sociologist, I think it’s one of the Ah, I don’t care. It’s okay, he has a specialty. Danbar who tells us that basically, you, beyond a certain number of connections, are not capable of apprehending social relationships with more than 150 individuals. Indeed, I had a bit of trouble making diagrams at that time.

Imagine that companies, as I told you earlier, they will arrive and perhaps some will never exceed 50 individuals. will remain at 1 like, I don’t really believe it.

50 individuals, it’s not at all the same thing to manage. It’s 150. Because there’s one thing that, this sentence, I wanted to say, it’s more or less balanced. I also wanted to convey messages.

Complexity kills companies. Complexity kills innovation, it kills autonomy. It’s a book you’ll find.

And unfortunately, the size makes things immediately more complex, but this complexity ultimately leads to death.

So, listen closely?

What makes the difference between someone starting out and someone more experienced is their ability to reproduce something, a recipe, let’s say, when faced with a type of problem. It’s the methodology, not the technique. Because I often tell developers, I think technical skills are great, but what really makes the difference in the end is your approach to the problem. It’s different from a process. A process is a kind of crutch that we apply to follow a recipe even when we haven’t understood it. And that changes everything because by repeatedly applying processes, we end up essentially reproducing things. You might know about cargo cults, a quick refresher, cargo cults are, it’s a student. which was studied in sociology, on a Guyanese people. This Oceanian people, was occupied during the war, World War II, and at that time, food and goods shipments were regularly made to their island. for the American soldiers, and they noticed that they always had a kind of ritual before receiving those famous crates of equipment. We had torches, made radio calls, that sort of thing. So they tried to reproduce it yesterday and it didn’t work.

Uh, I know the example is a bit far-fetched, but it’s supposed to make us question our ability to reproduce rituals without necessarily understanding why we do it. And I want to say, so agility typically, I’ve seen it practiced in many companies with exactly the same results, meaning no results, as people trying to bring crates home.

With that, there’s another topic, Mark Regan talks about it quite a bit. It’s uh what he calls ‘process people’. these people, basically the people in your company who always tell you, ‘it’s complex because it’s different, well, it’s different here.’ We’ve all heard that. Now, I don’t want to make enemies because I’m convinced that we all work, I’ve also worked on organizational stuff, I’ve also worked on processes that we don’t get from zero to that. But we still need to be aware that even if it’s done by very intelligent, very structured individuals who genuinely want people to, how can I say, do a bit better in the profession and so on and so forth. Despite everything, the process is because it will build the product.

Agility, typically, I’ve seen it practiced in quite a few companies, with exactly the same results, meaning no results, that people will try to bring crates home with them.

With that, there’s another subject, it’s Marty Cagan who talks about it quite a bit, it’s uh what he calls the process people. Uh these people are the ones who, globally, in your company, always tell you, uh it’s complex because it’s different, uh well, at our place, it’s different. I think we all know that. Now, I don’t want to make enemies because I’m convinced that we all work, I’ve also worked on audit, on organization, I’ve also worked on processes, we don’t start from zero, we know that. But, we must all be aware that it may have been done by very intelligent, very structured people who really want people to, how can I say, improve their craft and so on and so forth. Despite everything, the process is not what’s going to make the product. And it’s a characteristic of companies that have understood this, it’s precisely that we’ve introduced processes, processes, processes. What should alert us? Uh no, still safe. Anyway. What should alert us are the steering committees, uh the product meetings where we talk almost 50% exclusively about processes. I’ve experienced that, so uh I’m not saying uh I’ve experienced exactly that. When you start having product meetings, suddenly the time spent on the format, on how to make another process differently, on this and that, etcetera, etcetera. It’s because in fact you’re passing something, you’re passing next to something, you’re passing next to something.

And So, the accompaniment, it’s very good, I’m sure there are people who coach people in the room, it’s very good accompaniment. But the objective of the accompaniment, in the end, is supposed to be that we separate from each other anyway.

Uh, so when you’re on a permanent contract, for this type of position, I still have a bit of trouble understanding.

All that to say that, once again, we’re going to be confronted culturally with companies that are going to operate with a different structure and that are fundamentally going to go faster because there isn’t all that structure that’s going to annoy us. So we have to ask ourselves this question. And a second, ah yes, sorry, I wanted to tell you this. That, to a certain extent, I’ll just take a photo because I’m not going to, I’m not going to mention it. What we could be told, yes, okay, but you founded a company, it’s only 100 people, 700 people, companies with 10,000, 20,000, etcetera. So in reality, uh the processes you’re talking about, either you haven’t been convinced, but there are much bigger companies that need them, it’s normal. So I’ll let you take a picture of Steve Blank, Steve Jobs, Jeff Bezos, I could have mentioned others. There was one I had put in the list and since I found him unquotable, I removed him. But uh I think you all have his name in mind.

Uh, they all have a pretty strong opinion about processes as well. Not about it being unnecessary, but about it not having to take control of the company. It’s free, for once, it has nothing to do with AI, but it’s a message I want to convey.

Growth doesn’t kill products, it spoils them. This is another example I wanted to give you about the fact that there will be a fundamental difference between smaller teams and larger companies’ teams. You may have heard of l’agilification. gamification in French. I definitely prefer the French word for once. It’s not often that I prefer the uh how do you say, the French words. But here, for once, that’s the case. the theory that products are condemned to become rotten. Because there’s a financial incentive to do so. Because there’s a financial incentive to do it. Globally, most growths are done in the same way. At the beginning, you have a software on which you’re going to try to captivate users. You’re going to be more attractive in terms of price, you’re also going to aim very broadly in terms of functionality, you have a product that is super broad. And then gradually, what happens? You’re looking for profitability.

So you remove features that cost a little more to maintain, and then you increase prices, which means that, in fact, overall, in the end, you end up with products that are a little more you all saw Netflix at the moment with its prices, etcetera, etcetera. It’s normal, it’s an economic consequence of the model we operate on. Now, I think there’s a blind spot in this thing that doesn’t also take into account the size of companies. I’ve seen it in my company, and I’m convinced that it exists in others. That, yes, when we create specialized teams on a given subject, after a while, what happens is that these teams also need to justify their existence and they also need to shine, because that’s how we get advancement in a company, it’s not just by doing maintenance in an existing one or by removing features, even worse.

So, we quickly find ourselves with processes, with products that are much too complex for what they are. Because there’s also a question of ego, compared to the teams that also need to demonstrate what they are. So again, this is a cultural difference for me. We’re going to have, in the teams that have arrived, etcetera.

Another thing I want to put at the company impact level is the convergence of roles. what happens if suddenly we have devs who make interfaces? Because, well, that’s what happens.

What do we get if suddenly, I saw this at Alan not long ago, PMs start to make pull requests?

How do we discuss, how do we discuss a bit about these overlaps that will happen in the company? I imagine that’s part of the questions we can all ask, by the way. Is there one that will replace the other, are we in Battle Royale today?

And if there’s only one left, who is it?

There are companies that really ask themselves the question. And from the moment we ask ourselves this question, I think there’s no good answer, I’m telling you right away, I think there’s no good answer in the sense that you’ve seen Ratatouille? Ratatouille. Everyone can cook. not everyone will be a great chef, but a great chef can emerge from anywhere.

I think indeed that the skills that will be necessary to create a product are not dedicated to, are not specific to a given profession. Now, there are professions that will perhaps start with certain advantages because they have broader coverage over certain things, but it’s not a certainty.

What brings us to a question is, how are we going to recruit in the post-world?

Wouldn’t that be the end of the T-shirt profile? The T-shirt profile is a kind of unicorn that we’ve been chasing for 10, 15, 20 years, which is globally to say, we need people with a very general skill and this very general expertise. We’re going to complement it with a specific expertise. The general skill is to avoid having specialists who only fight with other specialists and who are more in a mode, since I know a hammer, all problems only look like nails. That’s all these situations, all these, all these situations.

Well, indeed, maybe tomorrow we’ll need people who have a general skill, always to avoid this famous syndrome. But who don’t need to have in-depth expertise in all fields. On the other hand, we need them to have a little bit of knowledge. So yesterday I was talking to Dimitri who was telling me about rake profiles.

Lots of little spikes like that, that we’re going to be able to amplify with a tool. so, let’s be precise, there have been a lot of scientific studies on the subject. AI can amplify existing knowledge. however, it won’t amplify something that doesn’t exist. Okay. well, yes, it can amplify the crap, the that we’re capable of producing because we don’t know a subject. But we’ll still need that little bit of knowledge. So there you go, that’s another change that’s going to happen to us and that will affect a lot of subjects in terms of recruitment. I’m obliged to speed up a bit because it’s 56. you knew I was going to mess up my thing. in short, we’re not going to talk about all that, fortunately, otherwise I wouldn’t tell you, we’re still here for two days. Uh societal impact, because we’ve seen all the impacts gradually, but okay, employment is starting to get disrupted, education, how do we train people? Uh ecology, obviously, geopolitics, who controls the tool, who controls the companies? it’s funny, we’re still coming back to Sauron and the ring. who, how do we run an economy with, companies that potentially employ fewer people, anyway, there are a lot of questions about that, but unfortunately, that’s not the subject today. However, what I want to tell you is that we are facing a new world that is coming at us. It’s not necessarily all negative, there are also opportunities.

So, certainly, as I told you earlier, I think it’s not the end of the world, I think there are still a lot of competitive advantages that we can capitalize on so as not to get eaten right away. Now, there’s still one thing to understand, it’s that certainly, these advantages exist. Now, the asymmetry of means, well, the asymmetry of resources, is still going to play a role. A company of 8,000 people and a company of, let’s say, 50, in the end, the cost structure not being the same, and if it’s not the same. There’s still an issue there. That is to say that, either we lower prices and become more attractive, or we set the same prices and make more margin, and make more investment. So, there is an economic equation that will still arise in front of all the companies that today do not adopt this way of doing things.

In a normal context, I want to tell you, I don’t think that most companies will submit to AI because, globally, it’s a cultural change and cultural changes are complicated to integrate into companies. In this specific case, I have a slight doubt. In this specific case, I tell myself that if the gains are there, I’m not claiming that it will be the case, maybe we’ll discover that in fact there is a big scam and that the bubble will explode in a year. Well, maybe. But if the gains are there, potentially we have a set of companies that are still going to slowly have quite aggressive competitors. And on which we will have to react. So we have opportunities. We have opportunities already because, on the one hand, there is still, like Dragos, no, with Playground, right? We have an opportunity because the geopolitical context wants us to have a lot of alternatives to building software in Europe vis-à-vis uh the US. So okay, that’s cool. Let’s launch ourselves, we have, we have the moment right now. And we have the technology for that.

We also have the opportunity, and I’m thinking of all the people today who work in software and who say to themselves, I actually constantly have my head in the handlebars, I’m constantly trying to think short-term. It’s complicated because my job is to work short-term. The only way, the only time I work long-term is defensively.

I’m constantly telling myself, okay, but is it going to hold the load? Is it going to do this? Is it going to do that? Okay.

But suddenly, we have a tool that allows us to save time if we know how to use it. reference, well, for previous slides. Which allows us to save time on strategy. We need to have people in tech, and when I say tech, I’m not just talking about devs, I’m talking about all the functions today that make the product. We need. of people in tech because we’re not making tech companies. And uh doing tech isn’t just writing code on a pad. We need people who step back, we need people who really think about tackling the problem and trying to find solutions for it. So for me, there are opportunities, we’re going to have to rethink the organization in the teams, we’re going to have to rethink everyone’s roles, remember earlier, convergence of roles, etcetera. It’s at that moment that we’re going to be able to seize.

One, we’re going to refocus on a job and not just on syntax. And two, well, we’re going to have, for all of us, opportunities within our companies to have an impact in a much more radical way. Or to create companies that will have an impact. So that, well, it’s now, and on the other hand, we must seize the moment. 18 hours. It was hot.

Well, thank you all. I think we don’t have time to answer questions directly, but at the same time, there’s no one behind, so there you go.

Uh, and otherwise, well, in fact, we can meet in the back room for those who have other questions, etcetera. Thank you, goodbye.

Let’s clap for him. Come on, over there then.