"The more AI can do, the more I want to do with it. I don’t fire people when I find a faster workflow. I find ten new things I couldn’t have tried before. " This is exactly what I tell investors that want AI to make us more efficient so I need fewer people.
Exactly the theory we emphasize always. it's about the building more values to enable the employability of more people that's the scale the world deserves. The FOMO strategy for the people with mal intentions not for the greater purpose
I think it’s important to make a distinction when you use the term “bubble” in your article. If you refer to AI use and future, clearly not only is not a bubble but it’s a technology revolution “just” started which will change radically the way we live.
However I feel that you are somehow mixing the above mentioned with economic bubbles. And when we approach this from an economic point of view to me there is no doubt this IS a bubble for many reasons: energy bottlenecks not be enough for data centres etc, huge investment in chips that will be obsolete before investment amortisation, circular investment of few companies and accounting manoeuvre. There are plenty of clear signs if one is looking for, one clear sign is Google issuing for the first time 100 years bond (last company doing that was Motorola at their pick before their fate changed).
You can have an economic bubble (and we will within the next few six months most likely) and at the same or after that also the technology at the centre of that bubble being a revolution. They are not mutually excluding.
Basically in your article I would have not used the word bubble which I associate mainly to stocks.
Said all that I enjoy your articles and knowledge sharing.
But who is calling for an AI bubble for 20 years? I think it’s something that came up in the last year for mounting logical arguments. It’s definitely popping before the end of year and most likely in the next few months. Stocks since January are showing signs of panic by falling daily a lot then the next day going up a lot. This is not business as usual. These are important signs that should be acknowledged. Yes, people can stay invested until bubble pops thinking they can time the market… but that is called gambling, not investing. (This is not financial advice. Do your own research before investing.)
You’ve nailed the distinction between the economic bubble and the tech revolution. This aligns perfectly with the 'Installation Phase' in cycle theory, where frenzied speculation funds the infrastructure (chips/energy) that often goes bust before it becomes useful.
The key to surviving that 'obsolescence' you mentioned is bridging the gap. Amazon survived not just because it was a 'tech revolution,' but because it was a dual-player. It acted as the Installation builder (infrastructure) that successfully transitioned to the Deployment utility (AWS/Logistics).
The companies that can span that gap, building the rails during the bubble and owning the utility after, are the ones that survive the amortization trap you described.
Tulips didn't become 100x more useful while getting cheaper. The product itself had no compounding value. AI does. So even if some companies are overvalued today (and some definitely are), the technology underneath is accelerating.
Precisely. The installation phase is expensive, and the speculative frenzy funds that necessary infrastructure. Much value is wiped out when the frenzy ends, but the generational wealth is built in the deployment phase that follows.
Amazon survived because it acted as a bridge..it was an installation company that successfully spanned to utility deployment. Those 'dual player' companies, builders of the infrastructure and enablers of the mass utility, are the only true unicorns.
I think this cycle is breaking the sequence. Usually deployment waits for installation to finish. This time, some companies are already scaling utility while the infrastructure frenzy is still running. If the phases overlap, the crash may not reset the board the same way. The companies already generating deployment-phase value aren’t playing the same game as the pure installation plays. Appreciate the dialogue here. I’m working through a lot of this thinking in my own writing on Substack, really grateful to have discovered this platform.
Enjoyed this. I’d only add: “AI gets cheaper” doesn’t settle the bubble question, because the bubble (if any) is usually valuations + capital allocation, not cost-per-token. Both can be true: a real platform shift and a frothy investment cycle with a shakeout. The practical takeaway still stands though — skill up, instrument your workflows, and measure cost per successful outcome (incl. review/rework), not per prompt.
I said it myself: the dot-com bubble popped. The revolution kept going. Both things happened.
Now here's where I disagree: "AI gets cheaper" absolutely matters for the bubble question. Because when a technology gets cheaper AND better at the same time, it creates more demand. Not less. That's the Jevons paradox.
That's demand catching up to capability. And demand is what justifies valuations.
Totally with you that cheaper + better can expand demand — Jevons is real. My only nuance is it answers “is there demand?” more than “are today’s valuations right?” Demand can surge while value capture gets competed away (or shifts up-stack to workflows, distribution, data). Dot-com taught us that too: the revolution was real; plenty of early “winners” still got wiped out.
So for me the bubble question isn’t cost-per-token so much as: who keeps margin once everyone has access? What do you think the durable moat is — distribution, proprietary data, regulated trust, or something else?
Also: the “sit it out” option isn’t neutral. In a Jevons world, delay can be the riskiest bet.
Not every AI company at today's price is a good buy. Some will get wiped out. That's always true. It was true with the internet, it'll be true here.
But "who keeps margin" is not the question most of my readers need to answer. They're not allocating capital into AI companies. They're just deciding whether to learn the tool or ignore it.
Totally fair — I hear you. For most readers it’s time allocation, not capital allocation.
My only point is: “where the moat sits” still matters indirectly because it tells you where to invest your learning (models vs workflow tools vs data + distribution). If you had to pick one lane for the next 3 months: promptcraft, automations/workflows, or domain-specific copilots?
Calling AI a bubble is emotionally comforting because it lets you keep your current workflow but bubbles don’t usually get better and cheaper at the same time. Even if valuations deflate, the capability curve keeps compounding, so the smart move is simple: treat AI like an asymmetric bet and build the skills/workflows now, worst case you get faster, best case you don’t get left behind.
Most people betting it's a bubble haven't thought through what happens if they're wrong.
If this isn't a bubble and the productivity gap between AI users and non-users keeps compounding the people who sat out 2025 and 2026 don't just fall behind. They can't catch up.
The asymmetric risk argument here is the one most people skip over. They're calling caution the safe bet. It's actually the riskiest one available.
What's the single thing you'd do differently tonight if you knew for certain it wasn't a bubble?
AI will drive economic growth - even if OpenAI, Anthropic or others fall by the wayside. Remember, the US had about 1.000 Car Brands in early 20th Century.
I don't doubt AI at all. Where the bubbles going to pop is when the Chinese models catch up. They are just a few months behind and have spent 1/9th of what the Americans have to get there. OPENAI would have to raise their prices 5x minimum to be profitable. Google will be the winner in America, Anthropic will be a niche player. OpenAI will be remembered as the most expensive startup in history. The Chinese take over third world and developing countries (most of the world's population) . Mitral takes Europe because they are GPDR compliant.
The asymmetric bet section should be the whole post.
That's the only argument that matters: If AI is a bubble and you learned it anyway, you lost some weekends. If AI is real and you waited, you lost a decade of compounding advantage.
One side risks time. The other risks relevance.
People love asking "is AI a bubble?" because the question itself is a stall tactic. It sounds smart. It buys time. It lets you keep doing nothing while feeling like you're being "thoughtful."
One scenario costs you some evenings. The other costs you your career trajectory. And that’s not even a 50/50 bet: one has almost no downside and the other has almost no ceiling.
The thoughtful people asking the question are making the riskiest bet in the room.
They just don't see it because it feels like standing still. And standing still feels safe. It's not.
Just from the little I've seen in the past year and a half, AI will be the most life altering technology since the last time we made it to this point in a not so distant civilization, and subsequently destroyed ourselves in short order after its arrival. This will inevitably and unfortunately be the result this time around as well. It's not if it will be used to destroy the world only when?
My son is a creative — artist, musician, film school grad — and he wants nothing more than to build a life in the film industry. He's also taken a hard stance against AI. In his eyes, it has devalued everything he pours his soul into. He's 23, idealistic, and genuinely believes that if he boycotts AI, he alone will defeat it.
I've told him — more times than he wants to hear — that he needs to learn it. Use it as an extension of his creativity, not a replacement for it. But any mention of AI shuts the conversation down. It's visceral for him.
I understand where his heart is. But it's hard for me to relate to his contempt for it, because I use AI for everything I do. I wake up looking for new ways it can help me improve — work faster, work better, just work. It's not a tool I dabble with. It's how I operate.
And that's what keeps me up at night. He won't lose his job to AI because it replaces what he does. He'll lose it because someone who does what he does also uses AI — and he refused to. He is, without question, the person who three years from now will wish he had started today.
The cruelest part? The creativity he's trying to protect is the exact thing that would make him exceptional with AI. He just can't see it yet.
Your son has taste. That's the part you should hold onto :)
Most people using AI have zero taste. They type a prompt, accept the first output, and call it creative work. Your son would never do that. That's his edge. He just doesn't know it yet.
And if it helps: the fact that he cares this much about his craft means he's already ahead of 90% of people using AI today. They have the tool but no taste. He has the taste but no tool.
The second he picks it up, it's over for everyone else, really.
The incentive shift is real. I don't need a team of 10 to launch something that looks professional anymore. I need taste + AI + 15 hours. Maybe even less.
Thank you Ruben. Very nice to engage as well. Always welcome to compare notes. I’m less interested in hype and more interested in designing for a more stabile future.
Exactly, they are dead ending trillions of dollars into AGI that can’t emerge because the systems don’t allow it. However, that system was built on a cell phone with two AI models in less then a month. If we don’t improve our systems, they will just keep offloading pressure onto people.
You talk about people not using AI and starting to learn. And you talk about those fully using AI. But what happens when all that changes because of AI? I'm still not sure where to position myself. Seems like no matter what I learn, I'll still be behind, still won't be enough as AI keeps improving. Wondering where we are really going, what I should really do. Give up and just get out my canvas and paint brush? That's not what I'm doing, but I wonder -- when I look back, what will I wish I had done with my time.
"The more AI can do, the more I want to do with it. I don’t fire people when I find a faster workflow. I find ten new things I couldn’t have tried before. " This is exactly what I tell investors that want AI to make us more efficient so I need fewer people.
Tell them. I integrated AI into everything I do and ended up hiring MORE people. 10x the ideas worth executing means you need more hands, not fewer.
Exactly the theory we emphasize always. it's about the building more values to enable the employability of more people that's the scale the world deserves. The FOMO strategy for the people with mal intentions not for the greater purpose
I think it’s important to make a distinction when you use the term “bubble” in your article. If you refer to AI use and future, clearly not only is not a bubble but it’s a technology revolution “just” started which will change radically the way we live.
However I feel that you are somehow mixing the above mentioned with economic bubbles. And when we approach this from an economic point of view to me there is no doubt this IS a bubble for many reasons: energy bottlenecks not be enough for data centres etc, huge investment in chips that will be obsolete before investment amortisation, circular investment of few companies and accounting manoeuvre. There are plenty of clear signs if one is looking for, one clear sign is Google issuing for the first time 100 years bond (last company doing that was Motorola at their pick before their fate changed).
You can have an economic bubble (and we will within the next few six months most likely) and at the same or after that also the technology at the centre of that bubble being a revolution. They are not mutually excluding.
Basically in your article I would have not used the word bubble which I associate mainly to stocks.
Said all that I enjoy your articles and knowledge sharing.
Calling for a bubble during 20 years (until it pops) makes someone very wrong.
Same vibe here. Yes if I call for a recession for the next 20 years, one year i’ll be right. But not being right on time is very wrong.
But who is calling for an AI bubble for 20 years? I think it’s something that came up in the last year for mounting logical arguments. It’s definitely popping before the end of year and most likely in the next few months. Stocks since January are showing signs of panic by falling daily a lot then the next day going up a lot. This is not business as usual. These are important signs that should be acknowledged. Yes, people can stay invested until bubble pops thinking they can time the market… but that is called gambling, not investing. (This is not financial advice. Do your own research before investing.)
You’ve nailed the distinction between the economic bubble and the tech revolution. This aligns perfectly with the 'Installation Phase' in cycle theory, where frenzied speculation funds the infrastructure (chips/energy) that often goes bust before it becomes useful.
The key to surviving that 'obsolescence' you mentioned is bridging the gap. Amazon survived not just because it was a 'tech revolution,' but because it was a dual-player. It acted as the Installation builder (infrastructure) that successfully transitioned to the Deployment utility (AWS/Logistics).
The companies that can span that gap, building the rails during the bubble and owning the utility after, are the ones that survive the amortization trap you described.
“Bubbles don’t get cheaper…”
You misunderstand. The tulip bubble isn’t the cost per token. It’s the value of the companies selling tokens.
Tulips didn't become 100x more useful while getting cheaper. The product itself had no compounding value. AI does. So even if some companies are overvalued today (and some definitely are), the technology underneath is accelerating.
Amazon lost 90% of its value after the dot-com crash. It was still Amazon.
Precisely. The installation phase is expensive, and the speculative frenzy funds that necessary infrastructure. Much value is wiped out when the frenzy ends, but the generational wealth is built in the deployment phase that follows.
Amazon survived because it acted as a bridge..it was an installation company that successfully spanned to utility deployment. Those 'dual player' companies, builders of the infrastructure and enablers of the mass utility, are the only true unicorns.
Much "perceived" value is wiped out. So we bet on the play, not on the individual actor(s).
I think this cycle is breaking the sequence. Usually deployment waits for installation to finish. This time, some companies are already scaling utility while the infrastructure frenzy is still running. If the phases overlap, the crash may not reset the board the same way. The companies already generating deployment-phase value aren’t playing the same game as the pure installation plays. Appreciate the dialogue here. I’m working through a lot of this thinking in my own writing on Substack, really grateful to have discovered this platform.
Spot on!
Glad to hear :)
Enjoyed this. I’d only add: “AI gets cheaper” doesn’t settle the bubble question, because the bubble (if any) is usually valuations + capital allocation, not cost-per-token. Both can be true: a real platform shift and a frothy investment cycle with a shakeout. The practical takeaway still stands though — skill up, instrument your workflows, and measure cost per successful outcome (incl. review/rework), not per prompt.
I said it myself: the dot-com bubble popped. The revolution kept going. Both things happened.
Now here's where I disagree: "AI gets cheaper" absolutely matters for the bubble question. Because when a technology gets cheaper AND better at the same time, it creates more demand. Not less. That's the Jevons paradox.
That's demand catching up to capability. And demand is what justifies valuations.
Totally with you that cheaper + better can expand demand — Jevons is real. My only nuance is it answers “is there demand?” more than “are today’s valuations right?” Demand can surge while value capture gets competed away (or shifts up-stack to workflows, distribution, data). Dot-com taught us that too: the revolution was real; plenty of early “winners” still got wiped out.
So for me the bubble question isn’t cost-per-token so much as: who keeps margin once everyone has access? What do you think the durable moat is — distribution, proprietary data, regulated trust, or something else?
Also: the “sit it out” option isn’t neutral. In a Jevons world, delay can be the riskiest bet.
Not every AI company at today's price is a good buy. Some will get wiped out. That's always true. It was true with the internet, it'll be true here.
But "who keeps margin" is not the question most of my readers need to answer. They're not allocating capital into AI companies. They're just deciding whether to learn the tool or ignore it.
Totally fair — I hear you. For most readers it’s time allocation, not capital allocation.
My only point is: “where the moat sits” still matters indirectly because it tells you where to invest your learning (models vs workflow tools vs data + distribution). If you had to pick one lane for the next 3 months: promptcraft, automations/workflows, or domain-specific copilots?
Calling AI a bubble is emotionally comforting because it lets you keep your current workflow but bubbles don’t usually get better and cheaper at the same time. Even if valuations deflate, the capability curve keeps compounding, so the smart move is simple: treat AI like an asymmetric bet and build the skills/workflows now, worst case you get faster, best case you don’t get left behind.
That's basically the whole article in one paragraph. You get it. Now go open Claude tonight
g00d aRticLe - jUst kEEp pLaYiNg ...
yep
Most people betting it's a bubble haven't thought through what happens if they're wrong.
If this isn't a bubble and the productivity gap between AI users and non-users keeps compounding the people who sat out 2025 and 2026 don't just fall behind. They can't catch up.
The asymmetric risk argument here is the one most people skip over. They're calling caution the safe bet. It's actually the riskiest one available.
What's the single thing you'd do differently tonight if you knew for certain it wasn't a bubble?
Yup, to wait and see is a bet. And it compounds against you.
I'd take the thing I spent the most time on this week at work, the task that ate hours, and I'd paste it into Claude tonight.
Much "perceived" value is wiped out. So we bet on the play, not on the individual actor(s).
What play are you betting on?
AI will drive economic growth - even if OpenAI, Anthropic or others fall by the wayside. Remember, the US had about 1.000 Car Brands in early 20th Century.
I don't doubt AI at all. Where the bubbles going to pop is when the Chinese models catch up. They are just a few months behind and have spent 1/9th of what the Americans have to get there. OPENAI would have to raise their prices 5x minimum to be profitable. Google will be the winner in America, Anthropic will be a niche player. OpenAI will be remembered as the most expensive startup in history. The Chinese take over third world and developing countries (most of the world's population) . Mitral takes Europe because they are GPDR compliant.
I see no one talk about this.
Who wins the AI race between OpenAI, Google, Anthropic, Mistral, or DeepSeek doesn't change anything for you and me.
It actually makes it better for us. More competition = cheaper tools = more access = more things you can do with it.
The asymmetric bet section should be the whole post.
That's the only argument that matters: If AI is a bubble and you learned it anyway, you lost some weekends. If AI is real and you waited, you lost a decade of compounding advantage.
One side risks time. The other risks relevance.
People love asking "is AI a bubble?" because the question itself is a stall tactic. It sounds smart. It buys time. It lets you keep doing nothing while feeling like you're being "thoughtful."
Can we afford to be wrong?
One scenario costs you some evenings. The other costs you your career trajectory. And that’s not even a 50/50 bet: one has almost no downside and the other has almost no ceiling.
The thoughtful people asking the question are making the riskiest bet in the room.
They just don't see it because it feels like standing still. And standing still feels safe. It's not.
That bet is known as Pascals wager.
Just from the little I've seen in the past year and a half, AI will be the most life altering technology since the last time we made it to this point in a not so distant civilization, and subsequently destroyed ourselves in short order after its arrival. This will inevitably and unfortunately be the result this time around as well. It's not if it will be used to destroy the world only when?
I don't write about whether AI will end the world. I write about how to use it on a Monday morning.
I focus on what I can control: whether I'm good enough with these tools to do better work, hire more people, and teach others to do the same.
My son is a creative — artist, musician, film school grad — and he wants nothing more than to build a life in the film industry. He's also taken a hard stance against AI. In his eyes, it has devalued everything he pours his soul into. He's 23, idealistic, and genuinely believes that if he boycotts AI, he alone will defeat it.
I've told him — more times than he wants to hear — that he needs to learn it. Use it as an extension of his creativity, not a replacement for it. But any mention of AI shuts the conversation down. It's visceral for him.
I understand where his heart is. But it's hard for me to relate to his contempt for it, because I use AI for everything I do. I wake up looking for new ways it can help me improve — work faster, work better, just work. It's not a tool I dabble with. It's how I operate.
And that's what keeps me up at night. He won't lose his job to AI because it replaces what he does. He'll lose it because someone who does what he does also uses AI — and he refused to. He is, without question, the person who three years from now will wish he had started today.
The cruelest part? The creativity he's trying to protect is the exact thing that would make him exceptional with AI. He just can't see it yet.
Your son has taste. That's the part you should hold onto :)
Most people using AI have zero taste. They type a prompt, accept the first output, and call it creative work. Your son would never do that. That's his edge. He just doesn't know it yet.
And if it helps: the fact that he cares this much about his craft means he's already ahead of 90% of people using AI today. They have the tool but no taste. He has the taste but no tool.
The second he picks it up, it's over for everyone else, really.
Very well structured thought process, which I have also been trying to explain to people!
thank you :) nothing good happens from watching from the sidelines, only when we open the tool.
This isn’t a bubble cycle.
It’s a compression event.
AI is collapsing the distance between idea and execution.
That changes incentives, not just markets.
The incentive shift is real. I don't need a team of 10 to launch something that looks professional anymore. I need taste + AI + 15 hours. Maybe even less.
Ruben, talk about this and a lot more weekly forecast. Always nice connecting with fellow thinkers.
And it’s nice having you in my comments, really.
Thank you Ruben. Very nice to engage as well. Always welcome to compare notes. I’m less interested in hype and more interested in designing for a more stabile future.
Exactly, they are dead ending trillions of dollars into AGI that can’t emerge because the systems don’t allow it. However, that system was built on a cell phone with two AI models in less then a month. If we don’t improve our systems, they will just keep offloading pressure onto people.
You talk about people not using AI and starting to learn. And you talk about those fully using AI. But what happens when all that changes because of AI? I'm still not sure where to position myself. Seems like no matter what I learn, I'll still be behind, still won't be enough as AI keeps improving. Wondering where we are really going, what I should really do. Give up and just get out my canvas and paint brush? That's not what I'm doing, but I wonder -- when I look back, what will I wish I had done with my time.
tonight, pick ONE thing you do at work that takes you an hour. just one.
paste it into Claude with this: "I do this every week and it takes me an hour. Show me a faster way"
you're exactly where everyone who's winning was six months ago. they just started.