Can you speak AI?

These are people who are capable of generating specific emotions in a group of people. And that part, unfortunately, that’s where it’s going to hallucinate, or it’s going to go completely off the rails, or it just won’t be able to. The fact is, it has no emotion, it doesn’t, you’re right, it’s more the technical side of marketing that gets handled, but not at all the substance of the message. Exactly.

And I’d even say, talking a lot with Vincent and with the clients we support in training residencies, or even just as a philosophy, it’s that our trades, the fact is, not everyone studies their trade inside and out, becoming generalists. A bookkeeper can be very curious and have side quests, love painting, love music, you know what I mean? Not everyone develops their curiosity the same way. But me, with the, and here I’m going to use my favorite word of the last three years again, ontology. But put very simply, it’s understanding your trade, understanding all the sub-tasks, the sub-intentions behind the fact that you chose this trade, because intentions today are more important than the tool you’re going to use. I was talking about it with Vincent.

Now, it’s more about becoming, you need to repeat that, intentions are more important than just the tool you have to pick. It’s, if you’re able to really clarify what you’re trying to do, what you’re trying to say, and you see the AI as a guide in that thinking, that’s where you accelerate. You know, velocity is direction and speed. AI gives you a speed that’s completely disarming.

You know, I can do in one hour what used to take me a month, literally. But the clearer you are, and here I’m bringing Vincent into the conversation, who’s an engineer, Vincent can do in one hour what used to take me a week, because now he applies the same principles I apply in design, he applies them in engineering. Vincent and I have become convinced that code, that design is no longer sacred.

It’s no longer a practice you can attach to yourself by saying, “I’m a designer.” So several things, really. And the fact is, you can detach from that role of being a selector, where all we do is choose: me, I’m good at Photoshop, me, I’m good at Figma, me, I. If you detach from that, and it’s hard, it took me 3 years, and 20 Vincents to talk it through. These are sacrifices.

We’ve put a lot of the time in our lives into learning things, but now there are things to unlearn. So give us, give us a concrete example, because that’s something that comes up a lot. You know, before 2023 we were in a WYSIWYG era, what you see is what you get, which was a principle applied to pretty much every tool, including even code development platforms or image, you’d manipulate things manually with your mouse to place things, or you’d type things with your keyboard. Now, I sort of declare what I want to see, what my intention is, and the AI is able to infer from what I wrote, because I want it to make it appear without me having to do the design process, without me having to learn to code. It shows me the final image of what it should be.

It predicts what it should look like. It doesn’t explain to me how it works, it doesn’t explain how it arrived at that result, but predictively, it shows me a final image based on how clear you are with me. It should look like this, the imagined version. And honestly, 95% of the time when you’re very clear, like Vincent says, when you know how to competently declare your intentions, describe what you want, why you want it to happen, that’s where the fascination begins, and that’s where you need to become capable of asking questions, of questioning yourself, above all.

I question myself every day now. Yeah. And without, without it being a philosophical introspection, it’s that the AI provokes that acceleration. And if we have no direction in our lives, then for sure it’s going to be a catastrophe.

Vincent and I found ourselves at a moment where we both had a direction, and we felt the speed accelerating, that we have to help people, we have to help people find their direction, because the speed is going to keep accelerating. And precisely, regarding your two intentions, because as you say, AI forces us to ask ourselves deep questions we didn’t need to ask before, because we had our little routines and everything was fine. Is this something that, for you, has come into your biz dev process, or in any case as a kind of service you want to offer with Crudle? The whole thing of supporting people through this, you know, is the idea to go down a layer below to go deeper? Well actually, at the start what we tried to sell was a platform, and what we realized trying to sell an AI platform, for us it’s more an infrastructure platform.

So it’s an infrastructure that we sell. Nobody buys that. Nobody buys that because, first of all, they don’t have the expertise. It’s very complex.

Which model am I going to choose? What are the costs that are going to be associated with that? How do I arrange things together to accomplish a result? It even goes as far as, how do I make sure my result is going to be reliable over time?

That is, that every time I use my AI-powered system, I’m going to get the right result, I don’t know, 100% of the time. 100% of the time. Honestly, I don’t know if that’s going to be achievable one day. 99%, 95%, 90%, so there’s a reliability rate that has to be defined, and that’s where we actually started offering services and supporting clients in making those good choices. Now, when I met Jonathan and saw what he was doing, that’s actually when we started grouping up with other disciplines, because the fact is, it’s more than a technological tool. It’s really an organizational transformation.

Mm-hmm. In that transition, there are several things. So there’s a technological side, there’s a strategic side, there’s a human side. And the human side is maybe the side that, with Jonathan, we touch a bit less now, or that we don’t support as well, but that we realize there’s a huge need for.

That is, there’s change management. Yeah. That the client has to do. And it’s normal, I mean, even me personally, when these tools arrived, I remember I told my co-founder, no AI is ever going to replace me for coding.

Never, it’ll never happen. And a year and a half, 2 years later, well actually, I don’t code anymore. It’s funny because you’ve almost defined our archetypes. You, you’d maybe be the technology expert, Jonathan strategy, and me human, like it was Artist Manager.

And the Talent Trap that I’ve been developing a lot right now, which is that you’re trapped in your own talent. Yeah. We talk about ontology and about words, but just to put a word on it, the fact that you say strategist. I think strategy is choosing the right moment to apply technology, design, or, as you were saying, human communication, you know, because for example, design is changing drastically.

There are people who are still attached to tools. Like when people use the word design, I say, first of all, design really isn’t a clear enough word. Even the word technology, I use technology too, you know what I mean? I think the infrastructure that Vincent has put in place lets me guarantee to my clients that they’re going to be able to observe results that are consistent and coherent with AI. Because AI hallucinates.

Me, I use AI to handle design, to replace my conception processes. Listen, not even 3 years ago, I was still making interface drawings by hand. And I’d give that to someone who interpreted my drawing, who redid it in Photoshop or Figma, who handed it back to a developer, who interpreted Figma, who redid it in code. And in all those iterations, which took a week, we lost quality, not quality on the conception, I’m talking about quality on the vision.

The vision got watered down more and more. And the developer had no idea about the vision or the initial intention, because he was in the code. The Figma guy thought he was the truth because he’d made a very beautiful interface, but one that was entirely disconnected from the final code. Now that’s over.

Now when I say something, the design system talks to the code, the code talks backward to the design, then the design comes back to the initial, to the initial intention, everything is connected. So what you just described, what you just described, is a kind of cutting out the intermediary, killing the middleman. Are you already able to model what the new global workflow is for, and by that I mean also modeling it for all businesses, because right now we’re all cutting out the middleman. It’s sort of the goal of Crudle, the goal of this discussion too, by the way, with Julien. It’s how we talk about this change, how we express visions and intentions now, how it’s connected to design, which is also completely being reinvented at the process level. Because if you’re able to declare the photo finish, or the final design of your experience, you can skip a month of work, of back and forth with a team of 5. It doesn’t remove the importance of talking to each other. Vincent and I talk more than ever.

So the relationship, no, but the relational side is super important, because since you have more time to think, well, you need people who are solid. He’s very solid on technology, indeed, and he asks me questions like, can you really go into production with this idea, Jonathan, you’ve got gaps, you’ve got security holes here, all that. AI isn’t going to settle the important conversations, but it’ll highlight them if you’ve done your work too fast or if you’re not competent. Me, if I hand something to Vincent, he goes very fast. You can’t put this into prod. So Claude, Claude is going to want to satisfy me, it’s going to say your product is finished.

Yeah, that’s it. It’s going to, it’s vanity. But precisely, Vincent, you work with people in a team. How do you manage the fact that people have to get on board with, or work with, AI now?

Do you have any resistance? Because there are, there are, I, there’s negotiation underway. I’d say there’s, we don’t have much resistance anymore now. There was some in the process of getting there. Mm-hmm.

It’s funny because us, at the start, we were developing decentralized peer-to-peer technologies. And actually, we never thought it would be our organization that would end up undergoing this transformation more than the technology. In reality, now our way of operating is that an individual has a project. So an individual has a result to deliver.

So it’s not even an objective anymore. That is, we define a metric, we say, here, you have to reach such-and-such metric. And each person is autonomous. So you’ve got peer-to-peer employees.

Exactly. We communicate together to exchange best practices, to give each other advice, sometimes to answer questions, but each employee is responsible for a result autonomously, and that employee, they work with AI agents that are in reality their team. You know, that was the, when Netflix bought the company I worked for, that was the thing that had won me over, it was the autonomy. You’re well paid and nobody tells you what to do, they give you the context and you figure it out. That, I had loved that.

In reality, it didn’t go like that, because we were a VFX shop and we weren’t, it wasn’t as easy as expected. But what you’re describing now is the, it’s actually the Netflix dream applied for real. Exactly. Exactly.

Because, first of all, if we think of the process we had before, before, we’d take a project, break it into tasks, assign responsibilities to different people. Today there’s no point in breaking it into tasks anymore, because a task gets done in 15 seconds. The fact is, we’re going to lose more time breaking up the work than doing the work. So we’re better off just understanding the context well, understanding the desired result well, as Jonathan was saying, understanding the intention well, and just doing it.

It goes much, it’s much faster. I’ve got like 20,000 ideas at the same time, guys. And I say come on board, but Vincent and I talk, it’s like, you know, you and I, Julien, we talk sometimes too, you know, sometimes you hear a conversation, you’ve got someone who’s autistic, ADHD, you know, there’s a lot of neurodivergence in the use of, the impression that I have is that I was punished and rejected for my ADHD my whole life, so I developed a shell, but when I arrived with the AI, it was like my best friend, it’s like it lets me follow my fractal thinking. I tell myself, if I upload a finished image of my interface, can I reverse-engineer all the way down to the code in a single shot? And I’d say, here’s an image I drew freehand.

Can you render it as if I’d made it in Photoshop, finished? Then it showed me. Can you do that in HTML? It did it for me in HTML.

Can you connect a CMS to that? But all those steps used to go in the other direction. Mm. Honestly, I did those steps I just named in one minute.

So there wasn’t really any philosophical or ethical reflection on, how many jobs did you eliminate in that minute, Jonathan? But I eliminated five. But I mean, we can think about the ethics of these choices we’re making right now later. Right now, today, we’re talking about the impact of velocity on our thinking.

Cognitively, I had moments where I lost the taste for my work, because I’d learned to get a lot of pleasure from, through the practice of drawing, through, for example Vincent doing code. There’s a pleasure there in the craft. OK. But now it’s as if I was watching someone who was in the process of learning my craft, and there I rediscovered the pleasure of teaching.

I rediscovered the pleasure of explaining why I did things. I rediscovered the pleasure of explaining to Vincent how I see design. What did I see that was missing in his craft? He saw the same thing on my side.

Jonathan, you can’t say that. Your prototype, it’s a month away from being in production. Here’s what you’re missing. There are a lot of people that Vincent and I realized matters a lot, it’s education.

Because right now, if everyone in the industry thinks they can do everything, that’s a big falsehood. OK. It’s going to hurt the industry if we don’t get into the surrounding discourse and say, look, there are still processes to put in place. There are still trades to, not preserve, but to describe well, because it’s not true that people are going to get replaced, and people have to transform themselves.

People have to ask themselves questions about why they do things. And it’s true that if you’re passive, yeah. And you don’t have your, I don’t know how to say it, the generalist concept. It’s not everyone who can become a generalist.

OK. But if you don’t develop your curiosity, for sure it’s going to hurt, because AI goes faster than us. One example, it learns from my image. The second time I come back into the system, it remembers the way I work.

It says, do you want to work like last time? Yes, I’ve got a new method for you. We could go faster. By the way, you could just give me a freehand drawing instead of giving me a fully developed Figma. So I tried that, you say, instead of developing a finished interface, I just made a freehand drawing, all crooked.

Honestly, try this at home, but you can just make a totally crooked drawing of a 5-year-old kid and the AI will bring it up to a very high-level interface. There’s no difference based on the level of quality you give the AI. It’s your intention that’s going to change the final outcome. I have the impression that we’ve gone from a world where the profiles that had the most prestige were the very calculating, mathematical, really meticulous profiles.

And now we’ve moved into a world where the world favors dyslexic, creative, a bit wild profiles. And because it’s no longer a big deal to know how to count. What counts is maybe having, making connections with things that others don’t see. What do you guys think about that?

Well actually, I think, I think it stays important to know how to count. There are maybe not as many people, on the other hand, who should be focusing on that. But you know, me, I’ve seen it in systems where we worked in the financial domain. It’s small, very subtle errors.

Mm. But it takes someone who knows how to count to spot them and make it possible to correct them. Now, I think the discipline that’s becoming more and more important is linguistics. That is, the choice of words is going to have a huge impact on the result, on the result you get.

And that comes back to the point about ontology, that is, sometimes in our vocabulary, we haven’t learned to choose words well, but there are subtleties in the meaning of words. In everyday life, words are, the two words mean the same thing. But if you look at the definition in the dictionary, there’s a difference between the two words. Learning to choose, or knowing in which context to choose this or that word, becomes extremely important with artificial intelligence. It lets me harp on this by saying, when you don’t know the difference between the word prototype and going into production, which is the case in, I’d say, 95% of the conversations I have with everyone, if I show them the product I now make in 1 hour instead of a month, and they look at the image, the interfaces, they think the product is finished, if I give that to Vincent, he still sees 3 months of work, an engineer who hears people right now say, hey, I did some vibing, hey, I did your project that would have taken a year in a week.

It’s not just insulting, it’s also very worrying. That is, people have to understand that the word prototype is not, it’s not the same word as a product in production. When I do design, I use words that Vincent doesn’t know. Say Vincent, he wants to make a little pop-up window appear at the top of the screen.

I’ve seen people try to describe that who didn’t know design. Well you know, when you click there’s a little window that appears, then it’s going to interpret that. It’s going to try to place it in its mental schema, saying, OK, a little window, that surely means a popup, or it surely means a modal window. If you’re a programmer who’s done this your whole life, you’re going to use the right word right away.

You’re going to say modal window. Nice size, progressive disclosure, second, because you want it to appear with a little fade in, fade out. Someone who doesn’t know design, they’re just going to describe a window that appears. Yeah.

And if you’re a filmmaker for example, you say, “I want to do a bird’s-eye view tracking shot, 3 seconds, fade to black.” You’re going to use the word, the cinematic language. Someone who’s a developer is going to say, “The camera moves fast above the city.” OK. But the more precise you are in the language, like Vincent said, the more varied your linguistics, the more you know, again, the more you’re relationally connected to other disciplines, the more your language is going to be nuanced, colorful, the more your solution is going to be deeply interesting.

You know, it’s the same thing as, yeah. What you just described to me is a little bit of what I wanted to learn the first time I really stepped into compositing. It was, how do I take what the director says in his emotional words, and transcribe it technologically with Nuke or a compositing software, you see. And that’s what fascinated me the most.

For example, you’ve got the boss who’d pass behind me and say, “Put a bit more green in the top right, a bit more red at the bottom.” I’d go OK. You know, I do 0.001, bam, it looks photorealistic. I go like, “What the fuck?” And what I have the impression you’re describing is a little bit that, it’s that precision in language, but that refers to an experience of the trade. I mean, I don’t want to steal time from anyone here, but you know, Vincent and I talked about the innovations that brought us to, OK.

Me, I didn’t study mathematics, but I’m a math nut, of the intersections. Vincent’s a guy who invents them. There are patents on some messed-up stuff. And you too, you’ve worked with things that were invented by the military.

You know, you were telling me you worked with Flame and the Flame system, you, a tracker, a bomb system, but one that was repurposed, and Vincent worked in games. They bought it on a military base, the Flame, in Luxembourg where I worked. It’s like, the tools I use today were used to do things other than good. OK.

What’s interesting is that Vincent and I still have a purpose of helping people with AI, helping people not to see it as a trap, as a threat, but rather an opportunity, actually, to rethink the world, to come out, as you were saying, Vincent, of the cave. We come out of Plato’s cave, we stop believing in the shadows on the walls and we ask real questions. That is, what can you really do with AI? OK.

And in design, me, when I met Vincent, we’re still in your conversations, Julien, and you’re part of them too. I try to separate the problems we can solve with AI so that people can get on board with us in the conversation. It’s impossible for someone to be able to be polymathic tomorrow morning if they were, sorry, across different disciplines. OK.

So you arrive at the, it’s as if Claude is a genius like Leonardo da Vinci. He knows 10 disciplines, knows way more than 10. He knows all the languages. He knows how to develop a product.

He knows how to optimize your car. He knows how to make your recipe. He knows like, you know, he knows everything. But the thing is, you have to know why you’re doing the thing in order to give a purpose to the AI, and not everyone develops that in their life. Developing your purpose, your introspection, it’s not, it’s not something you take the time to do naturally sometimes, it’s like you’re in your daily life, you’re in the operation of your company, me, I come in sometimes and people don’t have time to think.

So it’s fun, it’s fun to bring AI in to give them that back. I mean, Vincent and I, I swear, Vincent, I’m sharing this with you, for 2 or 3 months I was in a kind of internal solitude. I told myself, there must be other people who have this solitude in the face of what’s coming. And the conversations that come back to you right now, well, it’s not just conversations like, did you finish your PowerPoint, or did you finish your web page?

Those things are already finished. Yeah, that’s it. Vincent, he gives me challenges now. He says, “Don’t take a week, take an hour, Jonathan.” Show me what you can do in an hour.

What’s fun? It’s that it’s even better when he gives me less time. It’s not that the process has to take less time. It’s that Vincent says, “Can we move on to other things?

I’ve got other things to think about that are more important than you proving to me that you’re capable of pixel pushing. I really have other priorities right now. I’ve got four projects that have to be intentionally better refined, that you spend time in Photoshop. It’s fine.

The hour you spent, it’s fine. The client signed, he’s reassured. Let’s move on to more important things. That changes the game, being able to shift the conversation from execution to reflection.

Speaking of which, Vincent, can you tell us a little concretely what it’s about? What Jonathan is referring to? Well actually, you know, Jonathan brings up a good point, which is, you know, if we look a bit at the advances we’ve had, you know, the last time we had this kind of impact was about 100 years ago, and actually there were two technologies that had emerged at that moment, which are electricity and radio. Electricity brought a new form of speed that we hadn’t had before, and radio brought a form of communication.

What’s quite interesting is that we find ourselves 100 years later, and what are the two big innovations we’ve made in the last 20, 30 years? It’s the LLM, which is a new type of engine, a bit like electricity. By the way, it’s a bit like a commodity, you know. We buy tokens, we buy tokens and it produces a result, you know.

It’s a bit, it’s a bit like putting gas in a, wood in the, and on the other side, we have a new communication channel, which is the internet. So the combination of that offers new perspectives to humans that weren’t possible before these things arrived. So accelerating the daily tasks we can do, which bring very little value anyway, lets us reflect on what this transformation means for our societies. What does it mean for me as a human being?

What does it mean for my organization? So there’s all this reflection we have to do now to establish the new foundations of this era that’s beginning. But here I’m going to play devil’s advocate, because it’s not, isn’t it the case that certain people don’t want circles of thinkers to form, who would be more occupied doing repetitive, mind-numbing tasks, but now there are people who are becoming intelligent, who discuss among themselves and think about the future of the world? It’s not in the benefit of all the investor circles, it seems to me.

So, do you already see counter-attacks against this new possibility of thinking? Because me, I’m really excited, but I’m sure not everyone is excited by it. I think at the entrepreneurship level, there’s a lot of effervescence. You know, if we look at the period 100 years ago where we lived through the same thing, it was a bit, a bit similar. That is, there were people who’d had enormous success before that period.

And the people who had enormous success during that period, they’re people who weren’t around at all before. Mm. Yeah, but new players. Yeah.

But I don’t necessarily feel much resistance, you know, from those people. That is, first of all, they’re very involved actually in this transition. You know, if we look at, but still, these are new players, you know, because, I mean, we can look at Amazon, we can look at, but Amazon was born with the internet, which is one of the two big technological advances of the new era. But companies like that, where I think there’s the most resistance to change, and that’s where there’s probably a lot of reflection to do, but it’s at the level of our governments.

Yeah, it’s at the level of the governance of our societies and the laws that are in place that actually come to slow down the transformation. So we put in place laws to make sure the system was going to be inefficient. And no, I have to harp on this because I’m going to bring in the word geopolitics right away, because in the last few weeks, if people haven’t checked the news, everything Vincent just named is affecting things well beyond the applications we use and office-productivity systems, it’s affecting the structure of our societies. You know, just as, just as when I apply my governance methods to vision, intention, or the governance of design, or governance like Vincent of infrastructure, it goes out of the world of applications and rises ontologically all the way up to the level of societies, we do the same thing, the vision and the intention of a society.

And here, I’m going to give a pretty shortened example about what’s happening in the United States and in Iran. The acceleration of the world, the acceleration of geopolitics is also affected by the arrival of AI. If the United States, with Palantir, which is really a bit the Lord of the Rings story behind everything that’s happening, and they gave it a tool that unified their whole panopticon surveillance system, which was already very advanced. I mean, the United States has satellites that have been observing the earth in real time for about 20 years.

Our national aviation system is extremely advanced at the radar level. So they’re able to stick all this information together and know in real time, to the second, where a given individual is and which vehicle is where. In a way that was demonstrated, by the way, not long ago at the White House, there’s a video of the Chief AI Officer showing their system. Why does he show their system?

Well, first of all, it’s a show of force. The United States, it’s a propaganda machine for 100 years. They’ve dominated the planet with their power and their way of communicating. But there too, it’s that there was a very big risk that the United States would lose the race against China.

Mm. In this race to the militarization of AI. Me, I believe it’s done. I believe we’re witnessing this transition that’s important, that’s dangerous, but that’s also an indicator of how they see that they’re going to nationalize this innovation.

When you say you think it’s done, what do you mean? It’s that they’ve militarized the, the United States. OK. They’ve militarized their panopticon, they’re capable of intervening, of kidnapping a sitting president in place. They’re capable of intervening surgically, assisted by AI, I don’t know if it’s autonomous, what I mean is that narratively we’re witnessing the change of the world right now. If the United States hadn’t made that radical move, which I contest, which I really believe is an illegal war, but Iraq was an illegal war, you know what I mean, I’m not talking about geopolitics, but I’m talking at the level of the worldview, AI has entered our armies now, yeah. And, and they communicate a lot about it, but me, the Russians and the Chinese scare me even more because they say nothing.

Yeah, but the thing is, if again it’s fear that pushes us to lose, it’s dominance, we understand the power that the governance of AI gives to an army is enormous. So we’re not just witnessing the demonstration of what Vincent and I can do at the human scale, helping people in their company. We also see what it can do to societies. OK.

I believe there’s going to be a lot of diplomacy to do. Because it’s like a cognitive nuclear weapon, it’s that we can influence ideas, we can influence wars, we can influence a lot of other things than applications and corporate training interventions. Here, I, it’s, we go higher, if we look at the narrative of the world right now, it’s that we see where AI can take us on both sides. So me, as much as I was optimistic right now, I see the impacts I can have, Vincent, on industry, education, people’s imagination. As much as I see how much, if we let it go into the hands of the wrong people.

And here, I’ll just finish on this because I don’t want it to be a podcast about the end of the world, but at the, when they announced how they saw AI, they said it’s just an ontology. And the ontology we gave our AI is, the United States must win. OK. Yeah.

The United States, it’s the masters of the world. The AI that’s in Palantir has an ontological vision of the world that has a distorting prism on the hegemony of the United States. OK. It’s a bit in their constitution where they think they’re the chosen ones, right. We mustn’t forget that.

It’s that me, I reminded myself yesterday, the Americans think they’re the chosen people to guide the rest of the planet. But again, according to Alex Karp, and I’m not going to quote him more than one other time because there are opinions I don’t fully agree with, but he’s still the one at the head of Palantir, he’s still the one who put this in place with Claude, with OpenAI, he’s a master builder of this, OK, and he’s very articulate when he says, we’re no longer at 2 years out. We were at 2 days. OK, in the spectrum of the world.

OK, there are people who, for 100 years, have a, he calls it the nuclear time bomb, excuse me for the term, the, there’s a clock that’s tracked by scientists in the world for how many minutes we are from the end of the world. OK, it’s really a real thing. After the invention of nuclear war, we tracked that clock very closely. That is, what brings us closer to the end of the world?

OK. Why? Well, because we’re one decision away from making the wrong decision regarding AI, and so Vincent and I, when we talk like this with you, it’s not to scare people, it’s to tell them, become vigilant, become aware, use it properly, ask yourselves the right questions. It’s the same thing that has to happen in the military everywhere in the world right now.

Yeah. And take back, take back your sovereignty too, because me, I find that’s the very positive aspect of AI, it’s the aspect of, well, you no longer have a choice but to take back your sovereignty now, because otherwise you’re going to get wiped out big time. But the United States is demonstrating that, I just want to say it, they’re demonstrating it with force, maybe with too much radicalism. OK, because it’s in the hands of pretty radical people.

OK, but if there are, but I mean, the United States was already one of the biggest military forces in the world. Nothing has changed. They just added more precision. It hasn’t changed their attitude.

OK. But Palantir, when they said, us, we want to nationalize this innovation, that’s like going to the Moon, I mean, the Americans set foot on the ground and said, we are going to use it first, we’re going, they’re also the ones who dropped the biggest nuclear bomb on Japan, so that everyone watching this podcast doesn’t say, Chris is a, I’m not a nihilist, I’m just saying it’s the truth, it’s that the American military is putting in place a system that shows us the power of this. Yeah. The power of the, the power of the tools we use day to day.

Yeah. When we, Vincent, do you, because me, I often think, you know, I have the impression, the more I see the United States, the more I tell myself they’re a bit like unconscious teenagers who just want to blow everything up first before the others do it. Do you also observe a bit, regarding the Chinese, the Russians? I have the impression it’s a bit more mature.

Maybe I’m completely wrong, but is that the impression you have, or do you have a completely different reading of it? Listen, I think at that level it’s interesting to look a bit at history. OK. That is, if we look, if we go back to the last great era of change, it ended with 1929, a huge crash.

That depression was a deflationary depression. So there was a huge loss of value, and money actually became a very, very rare commodity. There have still been important changes that occurred in the last 100 years that would actually prevent the same type of crisis from happening today. The biggest of those changes is when the American government in the 1970s decided to remove the link that existed between holding gold and the American currency.

Now, what you have to understand is that, like say here in Canada, we have our own currency, but our own currency is actually simply a currency that’s pegged to the American currency. So our currency doesn’t really have value on the international scale, in a sense, I mean, by itself, you mean? Exactly. We’re able to go buy things. But you know, if we look for example at a barrel of oil, well, a barrel of oil is bought in US dollars, no matter who sells it, no matter who sells it. What we saw coming is still quite interesting, because we can link it to the narrative we’re hearing right now from the Americans.

We were talking about a climate crisis. We were talking about how we had to electrify our transport, electrify our various things, go toward clean energy. Now, who holds a large part of the production of that clean-energy equipment? It’s China. Mm-hmm.

So the biggest risk they had was that China would start demanding that those things be paid for in Chinese currency. Which would have hurt enormously, even us here in Canada, on our own currency. And then the narrative changed. That is, we’re now told that the climate crisis, well, it’s a bit of an invention, it’s not important after all, it’s not important, and then they come back to us with a discourse that’s more around democracy, it’s important, and AI is going to change, going to change our lives, going to change the world.

So then, well, what’s interesting in that? It’s, who holds, let’s say, the biggest models? The Americans. So with what are we going to buy models, we’re going to buy them with the American currency. Mm-hmm.

On the other side, what’s still quite interesting is that this technology needs chips. Yeah. And the type of chip, you know, Nvidia, Nvidia doesn’t produce any chips. Nvidia, Nvidia designs.

There’s one company in the world, not, there’s one that’s capable of producing this chip. OK. TSMC, Taiwan Semiconductor Manufacturing Company. And why is it interesting that it’s in Taiwan? Because actually China has an eye on Taiwan, and honestly I don’t think it’s, I don’t think it’s political, you know.

I don’t think it’s because they want the island. Mm. It’s because actually they want to have control over the chips. Yeah.

They want to be able to control the chips, and that’s really where we’re in a war. It’s not Iran, that’s a consequence of these tensions. But the real tension, that’s where it exists, because tomorrow morning, China takes possession of Taiwan, stops Taiwan’s exports to the United States, the American economy collapses. Because they no longer have the technology.

No, it’s that actually, since 2008, the United States has printed a lot, a lot of money. First, to cover the crisis, but after that we had the pandemic, all sorts of events that meant the Americans had to print enormous amounts of money. Now there’s a huge inequity right now between the very rich and the very poor. So the very rich ended up having a lot, a lot of money.

Even if the middle class doesn’t have more, the very rich saw themselves granted enormous amounts of money. And what did they decide to do? They decided to stockpile the money in things they thought would be able to keep their value over time, including a company like Nvidia. Mm.

Now, Nvidia is capitalized to the tune of 4.4 trillion. If ever China prevented Taiwan from exporting the chips and Nvidia ended up selling zero chips, mm-hmm. It’s a huge collapse of value. Again, the reason I brought up the subject you started with earlier, we could talk about geopolitics, is that me, in December, it had already been a year that I’d been working with Claude and certain tools that were connected to Claude, Claude Code was really starting to have dazzling performance, capacities to plan, to question me, I’d propose my methodologies and it supported me in improving my methodologies.

So there was really like a change in December in 20, you’ll be able to tell me about it, but I’d say 10 times more efficient than what I’d known. And then, when I saw Pete Hegseth, the American Department of War, who spoke up and said AI everywhere, no more restrictions, no more blah blah, you know, really radical, like, fuck the academies, fuck everyone who tells us no, us, we do it. OK, I went, oh my God, something’s going to happen that’s both good and difficult about this. It’s that the speed I thought I’d gotten used to multiplied tenfold every week. Now Vincent comes to me with new stuff.

We had the, the moment, I don’t know if it’s the right word Vincent, Open Claw, that few non-geek people maybe have seen go by, but autonomous agents, you know, stuff we can let work 24/7. There, I tell myself, “OK, if I have that, if I have access to that,” I was thinking of the microwave moment of the 70s, when the Soviets were accused by, by UNESCO of having used microwave weapons. They decommissioned the microwaves. But where did they end up?

The microwaves ended up in our homes. What’s in a microwave? There’s a magnetron. What’s a magnetron?

Well, it’s the thing that’s used to simulate black holes or make lasers. And the military kept the magnetrons to make lasers, communications, humanly responsible things. It’s very powerful, a magnetron, like an accelerator, don’t stand next to your microwave when it’s running, keep 3 meters even, it’s that, I mean, they gave us back Claude, me, I was like really happy, I’m happy they give us Claude, that we even still have the right to use it, but honestly, we talk about data sovereignty, and we’re just going to get out of geopolitics, in any case for my part, to come back to the word sovereignty, which is used by everyone right now, sovereignty, Law 25, bullshit, because right now, if the sovereign model is Claude, I guarantee you, those who use Claude with their files, think about it, it’s a model that’s American, under a lot of jurisdiction, if you haven’t followed the Pentagon story, I guarantee you they’re going to nationalize that model, and then France or Europe are going to have to react, their notion of sovereignty is going to have to maybe develop their own model, it’s expensive to develop a model. The Chinese, they’ve started to show that we could develop models with, they arrived with DeepSeek. We’re in for a ride, data sovereignty. I’d put that, I’m going to let Vincent weigh in on it.

The models, there’s going to be a nationalization of models in every country that seeks to keep its advancement. You, Vincent, are you also looking at developing local models that don’t necessarily run connected to the web, or not? You see, that, that’s a whole other expertise. What’s still good is that we have people in Canada, in Quebec, who are capable of working on that.

Mm-hmm. Me, I think, you know, in everything that’s the data-sovereignty aspect, there are still components we have, OK, today, that is, we have data centers. We have data centers that are still cutting-edge. Shout out to OVHcloud, which I interviewed recently.

Exactly. But we can also think of micro-logics, of scales. There we’re just talking, just talking, yeah, we’re just talking here, but you, across the whole of Canada, there are others too. Yeah.

Mm, the biggest challenges we have, because you know, that’s where, if we start comparing with, for example, AWS, GCP, Azure, we’re missing a huge number of tools to operate the infrastructure. That’s really where we have a big gap. Which prevents us a lot from developing, actually, from having our digital sovereignty, because we don’t have the tools to operate quickly. Do you think it comes from the fact that we don’t have venture capital in Canada compared to the United States? It doesn’t go fast enough because of that. I know I’m mixing the topics a bit, but I have the impression it’s linked to that.

The word risk, capital, maybe risk, Vincent, is there still venture capital in Canada? Well, I think so. After that, were companies encouraged to develop in those sectors? You know, me, I think if you look at what happened in the Valley, actually, there was a roadmap that was still pretty clear on the part of the people who deployed the capital since the 70s.

They really did it in stages. And every, let’s say, 10 years, there was a series of things that were done that bring us to today. Here, I think it maybe wasn’t as well structured as that. Mm.

OK. Yeah. That deployment of capital, what made it so that we end up with very, very few tools that let us operate quickly. Now, is it catastrophic? Maybe more because of, actually.

Ah, interesting. You, for your business, precisely for Crudle, did you, did you have investments? Did you go look for financing, or how did it go? I’m just curious.

Yeah. Yeah. We had, we had financing rounds, they were actually rounds we’d done before being in artificial intelligence. OK.

But we’re looking right now to do a new one to accelerate the deployment, actually. And will it be with American investors, Canadian, whatever? Canadian for now. OK.

I forgot the name, but the fund of, B, I’ve got the logo in my head, the one that’s like the biggest investor for tech companies here, the pension fund, no, something like that, no, actually us, we turned toward angel investors. Ah OK, there’s a network, there’s a network in Quebec called Anges Québec, which is a grouping of angel investors. People who believe in entrepreneurship, who believe in the next generation, they invest in all sorts of things, by the way, it’s not just in, it’s not just in technology. OK. They do support too.

So they have different services around that. So us, us, it’s angel investors who, who risked their capital because they believed precisely that we could make a difference, both in green technologies, but now they’re also very optimistic about what we do in artificial intelligence. Nice. By the way, on that subject, this afternoon, I’m going to go to the, there’s the Montreal Chamber of Commerce that’s organizing a themed day, it’s called Converge VT, convergence.

And it’s precisely about the ecological impact of tech. There’s the guy from Lufa who’s going to come speak. I’m eager to see, because if ever the price of diesel, of, of oil goes up, that we have to pay four times the price at our grocery store, it might be even more interesting to grow vegetables based on electricity in greenhouses here in Montreal. That could be fun.

I mean, let’s take advantage of it. Yesterday, there was, what’s his name, the Nobel Prize, Brassard, the Quebecer, I forgot his name. Anyway, it went through all the media outlets. 40 years ago, everyone thought he was crazy.

He’d observed things at the quantum level that could maybe enable cryptography and the exchange of information through photons. Yeah. Those ideas that were impractical 40 years ago, with AI, with the new developments in semiconductors, are maybe going to make revolutions. So ideas that are completely, I’d say far-fetched, become practical today. OK.

What’s his name? Brassard, his last name, it’s, I’m sorry, I forgot his name, his first name. Brassard, he was recognized yesterday, he’s a Quebecer. OK.

There are a lot of things that are going to come back. OK. There are a lot of far-fetched ideas that are going to come back, because mathematics, it’s mathematics. And I take the example of the thesis, when the inventor of Gemini, I forgot the name of the guy who, the head of Gemini, Vincent, if ever you remember it.

Anyway, he said we took a 150-page thesis with equations and diagrams on quantum physics. And we reverse-engineered the 150-page thesis with diagrams down to 500 words that described the thesis. And when we gave Gemini back the 500 words that described the thesis, the precise words, and Vincent will be able to add a little something on this. All the words in the 500 words are important.

There, you change a word, you won’t get the same result. They managed to recreate the thesis. So me, when I saw that as a poet, I said it’s fascinating. It means, it means we can reverse-engineer on the other side too, you know. It means if I show an image, you tell me what the prompt was to arrive at that.

But there, it explains, as Vincent was saying, the linguistics that has to be used. How does linguistics work? Well, it’s an equation. It means you can make an inference and reduce that equation even further.

It means those 500 words could become a single line. It means that those single words could, and maybe not just a word, but maybe. And there, I was like, in biblical language, there was first the Word, I can just tell you that there are people who use the grammatron. No but I mean, there are studies on this.

The Bible, it’s a grammatron, it’s, it’s mathematics. There are people who use AI to analyze the Bible. Incredible, it really has a lot of levels of reading. Yeah.

But, but in quantum physics, in electricity, in photonics, Vincent, there are a few things on this too in his team. Listen, it’s going, there are going to be revolutions that are quite substantially interesting that are going to balance what’s happening right now at the more, war and geopolitics level, but vaccines, research in materials, you’re going to have an era of renaissance. The Nvidia guy said we’re soon approaching a Country of Genius, because people are going to be accompanied by da Vinci, each having their da Vinci beside them to start imagining, maybe 6-year-old kids are going to make an innovation, my 16-year-old daughter is maybe going to invent a musical instrument. Ah no, but clearly, it’s that now we give everyone the chance to really dive into their imagination.

It can have a huge revolution on society, that, if we accept the challenge. But me actually, that, that makes me reflect on something I’d never thought of before. And you know, we’d already talked about it, Jonathan, that thing about Gemini, you know, but it’s maybe, it’s maybe a new way too of doing compression. For sure, we’re able to encapsulate a video or an image in a few words. Transferring a few words is much, much simpler than transferring the image or the video.

Yeah. As long as in the end you have the result and you have the thought process that’s the same, it can be, yeah, it’s really, no but think of Deep, of Open Claw, DeepSeek, they took a model, I think it was Claude or OpenAI, they used it to make something even faster and more effective, that’s it, right, I think, me, I’m not capable of doing, the thing is, Vincent does stuff I’m not capable of doing, but I do design today that, before, I used to dream of, like, ah, I’d love it that when I generate a website, everything is connected to all the origins of the documentation, that my PRD is never dead as a document, that the document that served to generate the software is always connected to the software, like that, it’s, look, it’s what, just for those who don’t know, the product requirement definition, OK, the spec sheet, if you go on the internet, all the big companies in this world, Facebook, OpenAI, Google, have an approach to defining a product. OK. Why?

Because it’s the ontology of how the product develops. Claude has a PRD that, that, and it’s an autonomous PRD. OK. That, that questions itself and that improves itself.

OK. DeepSeek too, Open Claw, I don’t know. But when I saw Vincent show me what we could do with Open Claw and say, and this is autonomous, it means it works on its own, learns skills. I mean, I have to get used myself, philosophically, to what that means.

Yeah, I haven’t even really finished. And my clients, when I introduce that to them, it’s too early because some of them aren’t even yet in the process of integrating the language that Claude can do things. There, I can tell them Claude, it can also do things when you don’t ask. Yeah, that’s it. It’s another game.

I’m not there yet myself, but I think it could really be the theme for the next show, precisely, autonomous agents and what’s the philosophy that goes with that? What do we, how do we have to update ourselves to understand how to work with that? Vincent’s going to have more things to say about that because he, you know, he develops them and he sees a bit what they’re about. Me, I use them, I use them and I see a bit what my users think when I put that in their hands.

It’s, it’s often philosophically the, the real problem is philosophical on my side. But Vincent, there are a lot of technical things to tell about this, I think. Yeah. Well, the first, the first aspect, it’s very pragmatic.

Me actually, I’m in the process of fixing all the annoying technical stuff I didn’t feel like doing. I hand that over because now you also have, what’s it called on Claude there? You have a thing called distributed, wait, is that it? Yeah.

Distribution in French. That is, you can pilot your Cowork from your phone, so your Mac, it’s going to access everything. It’s going to do all kinds of things, but you can just converse with your phone, and it’s like having a remote agent where you’d send messages to your assistant who’s on their computer at home. That, that’s already starting to be interesting on that side.

Like, for example, a really dumb thing, this morning, pragmatic, there was the Convergence schedule. I didn’t feel like putting that in my calendar. I just told it, here, here’s the internet link, put the events in my calendar with names that make sense relative to my objectives. Bam, it did that for me in 2 minutes.

Better than me going on the site, putting that in my calendar. It was super annoying, you see. So these are little things, but the more I think about it, the more I tell myself, “Ah, it’s like having an executive assistant available for your stuff.” We can, we can talk about it again next week, or the, the next, next, we can, I think there’s enough material for, for each week, you know. You see, even me, who plays a lot with this, I haven’t yet connected AI to my, to my own personal documents, or to my calendar, or to my email.

OK. I have that, I think it’s part of the reflection actually, it’s, to what extent do we want a productivity gain versus giving access to everything we know, to everything we know as a, as a person. Yeah. To an external model that’s not sovereign, for example, that’s not, so that’s where I think there’s a lot of reflection to do.

And you know, at Crudle, we’re in the process of developing an autonomous agent for ourselves that’s going to run on local models. Mm. And the main reason I’m doing it, it’s for me, it’s to be able to connect my email finally, and my documents, to a solution where I wouldn’t be afraid of what happens with, with my, my things. Yeah.

Me, the advantage is that I’m self-employed and I don’t have employees, so I have a bit less risk, but it’s clear that if you have a business with employees, I’d never do what I do, me. That’s for sure. I’m careful too with my clients because there’s Law 25 in Quebec that has a long arm, even, I’ve worked with clients who are still quite sensitive to that law, and other laws on traceability. As soon as you send a document into OpenAI or Claude right now, from your drive directly without doing any intervention on anonymization, all that, me, I see a lot of things, that we have interventions, Vincent and I, to make publicly, about, I call them the bros, but it’s not, it’s not mean when I use the word bro, it’s people who haven’t finished the reflection and who are already selling that to clients as if it were a miracle solution, Law 25, it has a long arm, because when we say traceability, I’ll just make my point, it’s that your document, if it passes through Claude and OpenAI directly, even the best of engineers, I bet, that, that, it won’t be able to trace where the information got lost.

There’s really an architecture to put in place. It’s not just design and vibing that’s going to settle that. There, it takes an engineer who checks the information. It’s put in a container.

That container is isolated. It doesn’t pass directly through Claude. We make a call to, we don’t make a call directly to Claude. Like, nobody is aware of that.

There it takes an architect. It takes engineering. So it’s not Claude and the others that’s going to settle that. Actually, them, they especially don’t want us to settle that.

Yeah, because we all become addicted to it and it’s easy. Me, I haven’t connected my emails either directly to an assistant, but I’ve already done a calendar, and it booked me some, some appointments. And what did it create with my clients? It’s that the clients would call me, I’d say, “It was really great, your, your presentation, I’m eager to see you tomorrow.” And there, in my spatial reaction, he knew it wasn’t me who’d booked it.

Right? What’s wrong? No no, it’s fine, it’s fine. But basically, it’s that it destabilized me that I had an agent, you know, I looked like someone who wasn’t aware of their own stuff.

Yeah, that’s why letting it book things for me, I’m not there yet. I’d tried Paulia, which is a kind of autonomous agent where you give it the link to your website and it’s going to analyze everything. And the goal is that actually it develops your revenue and they take a cut of 20% or 15% on what it earns. I sort of put monExpansion in there, and I stopped it right away.

It started sending LinkedIn messages to people, started wanting to invest in Messenger ads, everything. I was like, “Oh no no no, forget that. That’s maybe good for an idea that’s not connected to my brand, but I don’t want to risk my, my reputation, you see, with tools like that. But it’s worth talking about them, the tools that Vincent and I come across that are really interesting, and the ones that are a bit of smoke and mirrors that actually risk hurting the whole industry if we start just following the race and chasing the gold rush, because there are guardrails to put in place.

There’s really governance to put in place. How do you develop your intentions? How do you do your design process, and how do you do your infrastructure? It’s still the same three questions everyone should ask, because AI accelerates, you shouldn’t, you should define your direction on those three fronts.

Yeah, because I have the impression that the highest value in the AI era, it’s trust for a business. So if you, right now, well, for me that’s it again, that’s what makes me, Vincent quite a bit too, but we get attacked from all sides by proposals, and our clients too, like, you know, you can replace the client approach you spent 10 years building, and there your competitor arrives with a thing and wipes you out in a week. For sure it’s a fear to manage, but me, that’s it, it’s in the relational intelligence, it’s in having the courage for conversations, to name headcount as a real, important ethical subject, saying there are five people at my company, they all want to keep their jobs. It’s really philosophically important not to just arrive and say I can replace all that. Speaking of which, to finish, if we had to present what you do, Jonathan, and then we’ll move to Vincent, to someone who doesn’t know you, why would they hire you?

You see, if you had to sell yourself, say your elevator pitch, what would it be for those who don’t know? Yeah. Well actually, me, I organize residencies for companies, and I transform them into learning companies, as I call it. I teach them to unlearn the, the stigma of being tool selectors.

I teach them, it’s more important to choose tools. You’re going to, what do you let them solve as a problem, precisely? Well, I give them a methodology to do the governance of their project in this AI era. An approach to approach, what do I mean by defining intentions.

What does it mean to do design in the AI era? What does it mean to deploy an architecture, an infrastructure in the AI era? It’s really, I’m really a trainer, I’m a coach, you know what I mean? I take back the, the client-perspective side.

Him, what’s user-centric? What’s the problem he wants to solve when he calls you? They want to show an example. There’s someone who adopted AI at all three of those levels, and they want to see a project, they want to see a project in their organization that follows those three tiers. Declaring their vision, their intention, changing the design process, deploying into an infrastructure.

They want to see a prototype of that. They want to understand, like, how to talk about it, what’s the impact on the team, and what it costs. Getting on board with AI, but in the right way. Yeah, it’s like getting on board a train that’s going very fast, without a ticket. I give them, I give them a little seat beside me to let them see the view outside a bit.

After that, I, either I eject myself from the train because they’ve decided to stay in it, or, or they get off at the next station. These are projects that generally last from one month to a year. Generally, after a year, I meet a guy like Vincent for sure at the end of the train, because going into production takes a guy like Vincent who understands orchestrators, inferences, models, security, the laws. Vincent, the systems are observable.

It means they’re observable by a judge who’d say, prove to me that the model hasn’t stolen data, or prove to me that it’s good, secure, all that. It takes an engineer to do that. Me, I was a designer. So I design conversations about declaring your intentions.

I teach you to declare your intentions in the AI era. Be more explicit. If you say I want to do the audit of my brand, that’s not precise enough yet. What method are you going to take to do your audit?

What business landscape does your method touch? What are the metrics your method needs to be effective? Basically, what you do is you, you force the companies that want to launch into AI to clearly define their ontology, so that there’s no possible confusion. Exact.

After that, obviously that brings us to the 3rd tier, which is Vincent. Vincent, he takes my handoff, he has to start, yeah. And everyone is prepared. The IT people think like Vincent, the IT people look at me like a grain of sand in their chain.

When is this guy leaving? Perfect. Transition. Perfect transition so that Vincent can, exact.

I prepare, because Vincent goes at a crazy speed too in IT. So I prepare people in the language, I give them the right words. And when we get to the moment where the, I hand off the puck, I could say I get off the train, well, Vincent, there’s still a good foundation, the intentions are clear. The PRD is accessible for him too.

So you, me and him, we’re no longer in the, the code, the design, you know. We no longer have the, the drudgery of what we lived through in 2025, that is, there are steps, you know, we make a drawing, we make a design, we make, no, the language is established. So Vincent, when he arrives, there’s the, the spec sheets, he can finally do an IT acceleration job, you know, because me, I accelerate the design, Vincent accelerates the IT, he accelerates the infrastructure setup, the orchestration. So we see how you’re complementary.

In your words. Well actually, I think, you know, the solutions we use in AI, they’re mostly translators. It’s that actually we brought everything back to the same base. OK?

That is, before, each of us spoke in different languages. And now it’s called as-code. But you know, we have, it’s spec as code, design as code, security as code, architecture. It’s basically, everything is done in the same language, and we use AI to translate our intentions into that same language.

That’s actually what gives the great acceleration. That is, we no longer need to talk to each other in different languages, because what happened before is a bit as if us, well, we speak French but we don’t know English at all, and we have someone who speaks English and doesn’t know French at all, and we try to understand each other, to explain a project to each other, and at some point we end up recognizing a few words, like, ah OK, yeah, he means this, he means this when he talks about that, it’s a good analogy, but there, we just eliminated that. That is, we now all speak the same language through the machine, and the clearer we are in our intentions, the more faithful the translation is. After that, what we do is, with that code, it’s to see, OK, where does the client want to deploy it, how sensitive, for example, the data is.

What are the regulatory frameworks we have to face or that we have to think about. So that’s going to influence where we’re going to put, in our case, we’ve really developed lots of components, so we’re able to do a deployment in the cloud with, but we’re also able to do a local deployment, even down to a desktop, a desktop computer or a laptop. So we go across all those spectrums depending on, depending on what the client wants to accomplish. And after that, well, it’s a question of cost and reliability.

Mm-hmm. That is, there are different models we can use that each have different costs. The models also have expertise. That is, there are models that are better in finance, there are models that are better in legal, there are, so it’s choosing the, the right model, the right strength of model too, because, well, there are tasks where we don’t need, for example, to take Claude Opus, which is a model that’s still pretty expensive, or GPT Pro, GPT-5 Pro, which is very, very expensive, you know, even more expensive than Opus, GPT-5 Pro.

Yeah. OK. So there’s, there’s a balance to strike between the expected quality, or the complexity of the task, and the model we’re going to choose. And the last aspect is really reliability.

There, well, it’s always a question we ask our clients. If you want it to be reliable up to what percent do you want, to 60%, 70%, 80%, it can go up to, honestly, 100%, no, it can go up to 95. But that’s going to influence the amount of time we have to test, refine. Yeah, it’s like, it’s like VFX, 80%, everyone can do it. 98, it takes a pipeline like before, for big teams.

And after that, well, you know, if there isn’t really a big productivity gain in getting to 95, and that actually at 75% we’ve already automated the, the bulk of the work, it might not be worth paying for the extra 20%. Give us a concrete example maybe of a client or of a, of a specific case where you and Jonathan deployed something, just so we get a clear idea. Deployed. We’re, we’re not yet, we’re in the process of doing concepts together, of projects. We have five projects together. At the, the governance layer we’ve gone through, it’s the intention, the vision, and we’ve gotten to the design.

OK. We haven’t deployed yet. There’s one of my systems that Vincent investigated a lot to put into production. It’s what, what allowed us for almost a year to understand our two worlds.

OK. But it’s all going to be this year that it’s going to happen. We’re in prospecting to deploy that, so the types of clients, give me a list of, I don’t know, five typical clients you’d like to have or that you already, well, me, it’s going to be clients, good big clients. Vincent, I’ll let you weigh in, but me, it’s, I love education a lot, everything that touches on books and children’s education. The film industry, I come from cinema, helping someone with a pipeline in cinema, saving a Quebec industry, something like that.

I love working a lot with lawyers. It’s an industry that needs right now to, they’re very structured, giving them, it touches governance a lot too, the world of lawyers. So helping a firm, or firms that touch law. And what’s the, the billing model?

Do you take a cut on the money they earned, or do you bill for establishing the strategy? How much would it cost them, for example? What are the budget orders of magnitude? Well, me, it’s residencies that go from 10,000 dollars per month to 100,000 dollars per year.

OK. And Vincent taught me to no longer sell hours. So I sell, I sell on results now. One of the good things in my meeting Vincent, it’s that he modeled my entrepreneurial equation.

It’s like you, even you Julien, you were telling me, it’s, my value, it’s really the hour you spend with me, at your 100 hours, that is, it’s going to change your language forever. Vincent changed too how I think. I think I changed how he thinks. And that’s it, that’s our advantage.

Vincent and I, we’re in the process of clustering very, very interesting and diverse people right now to think about this coming year. And you, Vincent, what’s your business model? How do you bill, and really concretely, if ever we want to do business with you? How does it work?

Actually, there are really two sides. There’s the service side, so when we take an idea and bring it into production. So that’s code. We do everything at a fixed price, that is, there are a lot of risks in developing AI, and so us, we take that risk for the client.

For sure, us, we develop with the tools, you know, we have an approach you’d say is native. So we can, we can do a lot of code in little time. But it’s, everything that touches AI right now, it’s a lot of trial and error. Yeah.

Yeah. Which is the very nature of an LLM, in the end, right, it’s the, the statistics. So it’s logical on one side. So the more we can do iterations, the better it is. Yeah.

Our clients, I’d say we’re in pretty much every industry, because actually every industry needs to transform. Yeah. We have both the biggest companies and SMEs. Me, I’m particularly fond of SMEs.

I think helping them adopt AI is a mission that’s actually very important. Otherwise, we risk ending up with big, big groups that have too unfair an advantage over small companies. It risks putting the SMEs at risk. So, and to do that, well, listen, we do projects, it starts with Jonathan, we’re in the process of doing one, a project can start at 5,000, it’s very affordable, no, we’re not talking about 100,000, we’re not talking about 150,000, we sometimes do projects that are, but there, they’re hefty projects where we’re going to have, you know, 25 pipelines of, of stuff that talk together, you know, around fifty agents, all coordinated. You, I think you have, you’re more agnostic, you’re not just, because me, I see a lot of people who say, “Ah, I’m going to transform your enterprise,” and then they only talk about Microsoft Copilot, there’s nothing else. You, I have the impression it goes a lot further than that, that, since you’re a former CTO, someone who’s really hands-on in the code, you’re able to adapt to everything, not just proprietary systems where you’re forced to pay a subscription because you already have those tools in place, am I wrong, or exactly?

And actually that comes to the second side. That is, after that, what we sell is that we’re an LLM distributor. So we have lots and lots and lots, and we resell tokens. OK, to people.

Interesting, I’m going to allow myself to just refine my, my answer. So when I make prototypes to bring them toward Vincent, it’s between 10,000 per month and 100,000 per year, because it’s 10,000 per month. You have a prototype that runs, and you have a finished product. Yeah.

So that’s just an understanding of what I used to do in 1 year, in one month. But if you want to hire me for 1 hour, I have lots of packages on my site, ParaCost, that let people say, “OK, I need to reassure myself with my team about how I could transform my company with AI. I don’t yet have money to pay a guy like Vincent or you, but can you give me a bit of a roadmap?” I do roadmaps with my method.

In 1 hour, you’re going to know what, what the maturity of your company is, and your organizational capacity to use, again, to develop those capacities. It’s not true that everyone is mature in the same place. That’s quite fine. You know, sometimes we look at clients and we put that into our system.

You know, Vincent and I, we have our systems now to analyze in one minute whether the client is ready or not. There, just by listening to them talk. I don’t need to spend a week of analysis. I listen.

What’s he telling me? I listen to his brief, the words he uses, and I say, “Ah man, 60% chance it works.” You guys are a bit like, you know, a good, a good accountant, you pay him but he’s going to make you earn a lot of cash, because he’s going to stop you from doing a lot of dumb things and spending for nothing. It’s an investment. Yeah, it’s insurance on the future investment.

And Vincent, he takes very proactive approaches on that too. We take the risks for our clients. Me, I’ve exposed my method, I give it away. I’m very generous, I think, in how I give it.

I teach my method, because I think me, I have 64 types of questions. If you answer, if you answer five of them, it’s going to give you a score. I can tell you where you’re going to crash. By the way, the presentation you gave at EAI was really well received regarding that.

Yeah, I, I had three people who took the method, they made their product, and their concepts came to present it to me and Vincent. Some concepts incredible, honestly, some others incredible too, but problematic. Because again, people’s perception, when they live conception like this for the first time, they think it’s finished. In reality, it’s just beginning.

Nice. Well, that’s perfect. And, well, on my side, monExpansion, me, it’s really the, the human aspect, people who think that with AI they’re kind of, stuck, but with or without, actually, that they’re people who are very, very senior in their field, who are well paid, and who think they can no longer change their lives, that the excitement of youth is over, it’s over. Well, me actually, it’s the opposite.

I tell them, thanks to AI, but not only. It’s also thanks to coaching and to personal-development techniques. Actually, it’s the best moment in a life, especially if you’re 40, 50 today, to take risks again. Money, it’s just money, it goes, it comes.

You have to do it responsibly, but precisely, I support people in taking the taste, in taking risks again, because I think that today, if you don’t want to take risks, you really put yourself in danger. Ha! Strangely. So I find that we cover, we cover well, you see, we’re a bit like the three musketeers who cover well all the aspects of the transition, so that we don’t fall into a spiral of the end of the world.

Quite the contrary, we take action and we propose pragmatic approaches. So me, it mattered to me too, to do this show with you, because precisely, we’re solution-oriented. Of course, we’re lucid about the danger, about the dark things happening around this, but as much, as much, to bring a constructive or positive energy into it, so that, well, so that we use it for good things. So, a final word from each of you, because we’re going to keep this editable anyway.

Me, I can go. I’d say the important thing is, start now, because it’s, usage gives meaning. So don’t stay too much in your, your fiefdoms or your fears. And even beyond developing your curiosity, I mean, it’s really, there’s a, a transformation of our cognition happening right now in society because of this. So use it and ask, honestly, it’s like the ask five questions, the five whys.

Ask yourself the same question five times differently to really see what AI, how to explore the uses of AI, that’s more important. The meaning is going to come after. For sure, if you want to understand how it works, that’s another, more scientific approach. I have nothing against that.

You can go to university, study mathematics. OK. But poetry, it’s through usage. Build poems, ask yourself questions, the meaning comes after.

Magnificent. Well, listen, see you next week or in two weeks, we’ll see depending on your schedules, to talk about autonomous agents and how much we have to give them control or not. People have questions. I imagine that people, I mean, we covered AI today, true to ourselves.

So if there are questions you see flying around on the web, Vincent and I get, Vincent these days gets a lot of questions. So I’d be curious, precisely, where do people contact you? It’s LinkedIn mainly. Yeah.

OK. Well, listen, I’m going to put the link below the YouTube video too, but of course, we’re going to publish this on the networks, on your own respective networks, and people will already be on your account, so it’ll be fine. Thank you very much, Julien. Thanks.