You might think that AI is useless, dangerous, that nobody wants this slop written by clankers and that this whole thing is a ridiculous bubble which will only burst into flames just like crypto and the metaverse did, only with much more real consequences due to the amounts involved. If that’s the case, you’re like Steve Ballmer predicting the iPhone will tank: a short-sighted dinosaur unable to see the comet coming to their doom.

Yet if you think that AI is the future, that developers already are experiencing huge performance boosts, GPT-5 is coming to replace most white-collar jobs and that AGI will resolve all of Humankind’s problems by 2030 then the question becomes: how much nutritional value do you think Sam Altman’s bullshit contains exactly? Are you so gullible to think he’s waving his superintelligence flag in every media outlet for another reason than hiding the fact OpenAI completely stopped progressing?

The illusion of intelligence

Now that everyone feels offended, it is time to explain what LLMs—the technology behind ChatGPT, Claude and others—really are. They are made to be translation systems. French to English. Long text to short text. Soft information into structured data. Functional specification to code. Those are the key capabilities of a LLM.

Systems like GPT-2 or BERT were already pretty impressive on their own. When GPT-3 was released, it was the first time that we started seeing AI companies saying “this model is so powerful we don’t want to release it to the public yet”. And sure enough, the capabilities were amazing. From general knowledge to translation it felt to specialists like this opened the door for countless applications.

Yet what did the trick was GPT-3.5, also known as ChatGPT. Because you see, foundation models are not trained to work in any particular way, they just “hold” a representation of the world and its translation into text. ChatGPT invented putting it into a chat format.

Suddenly, you get a chat application that can answer any question on any topic. It was indeed revolutionary. Because it really feels like you are talking with a human. At least, until you start digging. Re-hashing Wikipedia works just fine, but specific knowledge starts being thin and reasoning falls apart completely. Any subject matter expert deep-diving into content written by a LLM—even the today’s best ones—will tell you that it is absolute garbage.

As a matter of fact, chat is a terrible use-case for LLMs. As stated before, they are dumb, inanimate translation systems. They appear to be human-like when you inject billions of dollars in their training, but the illusion falls apart quickly when you start poking.

The Capacitor Effect: why LLMs peaked already

Proud owners of EVs—or anyone with a phone and some level of observation skill—will have noticed that the closer you get to 100% the slower the charge goes. In fact, you will be shocked to learn that both your car and your phone are lying to you: it is impossible to charge them to 100%. Imagine all little electrons going into a bar. At first the bar is empty so it’s easy to get in. But the more packed it goes, the longer it takes for little Timmy the electron to find a spot on which to stand. Up to a point where there is still some space left but you really need to wait for the crowd to get into a special position to form the space for just one more.

The same thing happens with LLMs and their benchmarks. At the beginning, we were having a huge room for improvement. But the more billions were poured into compute power the more the easy problems got solved while the more complex ones remained elusive. This can very easily be observed when you graph the performance of LLMs on benchmarks, the progression has slowed to a halt for the past year.

Everyone who went through university knows what this is about. Most improvements in performance that you see these days are actually due to the fine-tuning method. Specifically, they are trained directly on the benchmarks. It’s exactly like taking an exam and training on the specific question types that you’re going to get instead of developing a deep understanding of the matter.

The fact that LLMs reached their current level of capability can be seen as a miracle. Knowing how they work under the hood, there was absolutely no reason to think that spending $10 million on training GPT-3 was going to yield any result at all. And yet they are insufficient on their own to achieve any kind of human-like intelligence. So it can also be seen as a malediction: all that money spent on marginal gains while actual research could have been done.

For the better or for the worse, current models are however improving in one particular area: their cost. While heavily gaming benchmarks, obviously, latest models display a great level of usefulness at much lower cost1. Soon we can expect that LLMs running on our phones will have the highest level of useful capability that a LLM can have.

Sam Altman: not a clown, the whole billion-dollar circus

Our little brains, whose baseline reasoning capability is much closer to LLMs than we’d like to admit, tend to imagine that if from GPT-3 to GPT-4 we’ve seen such a huge boost then the increment that GPT-5 will deliver will shatter the foundations of human society and GPT-6 will be indistinguishable from God.

Obviously, this will not be the case. But little Sam has gained a lot of traction, now sitting at the highest tables within the US elites and received unfathomable amounts of funding—at least in theory. He promised to deliver AGI and now needs to keep people believing because as soon as they stop the whole thing is going to come crashing on his face.

So yes. You’ll see Sam Altman in every media outlet claiming that AI will wipe out entire job categories or that making an AI startup is akin to building a nuclear bomb. Honestly every single one of his public interventions are hilarious. Because as we’ve covered so far, nothing can be further from the truth for at least the next 10 years.

Yann LeCun says LLMs are not going any further and that we need more fundamental research to get closer to human-like intelligence. Transformers—the technology that made LLMs possible—got discovered in 2017 and took 5 years to pan out. They have been decades in the making. How long more do we need before another breakthrough?

On the other hand, OpenAI’s costs are absolutely above the charts. And so are the costs of the rest of the industry. And so are the $500B promised for the project Stargate. That is a lot of money set to burn in order either scale something that maxed out or train a technology that doesn’t exist yet.

And that’s not all. There is an insane talent war going on where sign-in bonuses are rumoured to surpass $100M. In fact, most of OpenAI’s talents got poached by Anthropic, which probably explains why lately Claude has taken such a lead. In fact, if you look at OpenRouter’s data2, it appears that OpenAI is absolutely sidelined:

Last week, OpenAI had only 6.2% of the market shares. They are absolutely dwarfed by Google, Anthropic and DeepSeek that total ¾ of the market together. That’s how you understand the deception of Sam Altman: OpenAI, the poster boy for the whole industry, is barely relevant to business cases today. They could disappear tomorrow and it would have no impact.

The NVIDIA tax

Talking of talent war, NVIDIA is competing against itself, with its stock so high that their talents just quit out of being so rich. They are at the heart of why AI is so expensive, and beyond the ridiculous pay of top AI engineers most of everyone’s money just goes to them.

You see, AI is essentially a machine that computes every possible connection between every single one of the million words you can feed it at a time and decides which one is more likely to have meaning. This process is extremely compute-intensive and requires very powerful, dedicated hardware.

It is estimated that roughly half of this hardware is used to train models (feed them every single text ever written until they learn something) and the other half is used for inference (answer to actual queries).

The training process is very expensive, but also is done in several steps. The first and most expensive one is the creation of the foundation model. Presumably, given the plateau mentioned above, there will be no need anymore for such step fairly soon. Or at least not nearly as much as today.

Then what about inference? GPUs are made for games and 3D. The actual thing that you need is a chip like Google’s TPU, Groq (with a “q”)’s LPU or Cerebras, which are much more efficient. To compare what is comparable, Cerebras is able to serve open-source models at ten times the speed of competition. Meaning that while huge leaps in LLMs are not going to happen just now, huge leaps in the hardware and efficiency are still going to arrive pretty fast.

And to add insult to the injury, today, a Bluetooth-connected backdoor toy has more computing power than all of NASA during the Apollo missions. The destiny of compute is and always has been on edge. There will be no exception here, AI will not be computed in datacenters. iPhones already come with AI models running locally, it’s only a matter of time before top-of-line models are running on every single consumer device.

To summarize, right now everyone is buying unholy quantities of GPUs from NVIDIA, but from those only half is actually used for business needs, and when dedicated chips are mass-produced you can slash that by a factor 103, and soon enough you won’t even need any of that because most of it will happen on consumer devices. If you think that your car’s value takes a hit when you take it out of the dealership, try buying a GPU in 2025.

Faster horses, faster humans

As said Albert Einstein, when asked what would be the future of transportation, a gentleman from the 18th century would certainly say they want faster horses. The same applies for every single technology. Just have a look at futuristic 80s movies, their vision of today’s technology is absolutely ridiculous. Cars would be flying but you would still be getting your taxi at a stand?

Now AI has human-like qualities, are they going to overtake white-collar jobs? Pinocchio says AI will replace normal people, but is already diluting his wine between 2023 and 2025 where the tone switched from “some jobs will be 50% AI” to “hey look we found some jobs where AI helps a bit”.

The jobs to be replaced according to the 2023 paper were:

• Interpreters and Translators

• Survey Researchers

• Poets, Lyricists and Creative Writers

• Public Relations Specialists

• Writers and Authors

• Tax Preparers

• Web and Digital Interface Designers

• Mathematicians

• Blockchain Engineers (is that even a thing?)

• Court Reporters and Captioners

• Proofreaders and Copy Markers

• Correspondence Clerks

• Accountants and Auditors

• News Analyst, Reporters, Journalists

• Legal Secretaries

We are two years later. Did anything happen to these jobs?

LLMs being a translation tool, a third of translators are losing their job to AI, alongside with a quarter of illustrators. Working for a content website doesn’t seem to be a very safe career either, although this might have less to do with AI and more to do with the general context of content proliferation. There could be one or two other items in the list truly affected, but other than that it seems to be business as usual.

Is there any use for AI beyond translation then? We can once again use OpenRouter to understand what the majority of AI services are being used for.

• Code assistance tools

• “Role-playing” models (read: AI girlfriends and sex bots)

As it turns out, Melon Tusk got officially crowned the Incel King for breaking the Japanese Internet with an AI girlfriend. Host(ess) clubs are a real thing, in particular in Japan, and so are “dating sim” or even Maid Cafés. The need for companionship is extremely strong for some parts of the population, but solutions to that problem require either workers undergoing tremendous sacrifices or extremely shallow solutions. AI on the other hand delivers a fantastic blend—whether you like it or not—of being personalized, patient, conciliatory, educated, always available, always willing and all that for a very decent price.

Whatever is the explanation for humans still doing essentially most tasks—whether it’s insufficient quality from LLMs or lack of liability—the fact remains that AI ended up shining first in a field where humans do not. You need to expect seeing that happening countless amounts of times in the future: AI is simply good at different tasks than humans4.

Not a bit of productivity found

The big flaw in this line of argumentation comes from coding tools, which appear to occupy the vast majority of consumed tokens, with companies like Cursor or Windsurf flaunting billion-dollar valuations. But do they actually make developers more productive?

Productivity measuring tools tried to answer just that. In a study published by Uplevel, they couldn’t find any evidence of changes in productivity for developers after they started using GitHub Copilot. Which is consistent with this author’s own findings: while you can experience the occasional slam dunk, AI will usually only act as a friendlier alternative to the highly pedantic StackOverflow. In other words, it is a great learning tool, but a poor employee.

Worse than that, if used incorrectly it can blow up exponentially. Increasing number of candidates are using it in their tests, ending up with code they hardly understand themselves, let alone are able to defend in an interview—often because it is so stupid that you cannot defend it. Developers start taking the eyes off the road and end up writing 5x too much code in 5x too much time. Vibe-coding is even suspected of being behind major data leaks (although probably not yet). Coding agents are actually 10x juniors: same skill level, 10x the potential for harm. This is going to hurt the development industry so bad.

What you need to understand is that a developer spends only a small amount of their time doing what AI can do: transforming a human-language specification into actual computer code. According to Microsoft, this represents about 44% of a developer’s time. Inside of that, roughly 50% of the time is spent reviewing your own work, making sure it fits the specifications. Now let’s say that AI boosts this by 30%. You’re down with:

That’s 7% of developer time saved by AI. When being overly optimistic. Weehee.

This might come as a surprise, but even the development industry which is hailed as the use-case for AI fails to show any kind of productivity gains for it. AI might be a great help, as a teacher for example, but if it brings any productivity at all the gains are so small that they cannot be measured.

DinoSquad to the rescue

As a CEO, even if your business is shampoo, you are going to feel an itch to make a bold claim about your company being AI-first in an attempt to stay relevant. At this point most big players fell into the trap, and we’ll use Salesforce to exemplify this.

Salesforce is a complex set of products. Essentially they have one “offer” for each bullet point that you could ever encounter in a board meeting, made accessible in the form of a huge license fee which may or may not give you access to some piece of technology that solves said problem. Behind that you have huge IT integrators, usually outsourced in Asia, which take months to move a paperclip with setup fees matching up the licenses.

We’re not here to understand how exactly they manage to convince their customers to spend so much money on something that should be a tenth of the price, but the fact is that this whole machine works because the very point of Salesforce is being complex, opaque and hard to use.

So when they say that they are going to have Agentforce, supposedly a generic and simple solution to be plugged on top of a platform whose core culture is custom integrations, it appears that the market doesn’t like it too much and the share value drops by 20% over 2025, without a single success story of Agentforce in sight (outside of the Salesforce website, obviously).

Essentially it’s what always happened. New technologies are not embraced by old players to increase their business efficiency by a marginal amount. Especially not when the industry has turned into a gridlocked field designed to prevent new entrants from coming in. The change always happens by making the industry irrelevant in a first place.

Which also translates in the workforce. The skills required today to be a Salesforce employee do not match up the skills required to run a successful AI business today. There is no training in existence that can prepare you for this, simply because no one knows yet what even are the job descriptions that you need to work on AI. Prompt engineer? Agent developer? No one knows!

At this point, the nascent AI industry—which is the one that builds on top of LLMs—still needs to take shape. This will happen through a darwinian process, where the people who made the right bets now will reap the rewards in 5 to 10 years.

Saturn’s revolution takes 10 years

There is no telling how much, by lack of tools, developers have influenced UX patterns—and even entire workflows—over the past decades, but you should consider that everything you know and use daily is up for grabs.

The Web was launched in 1989. By 2000 it caused a market crash. In 2004 was the launch of Gmail which introduced the “Web 2.0”. Essentially, that was the first application built for the Web realizing the full power of the platform. It took up another decade to build the frameworks and tools that allow the Web we know today to exist. Millenials have always known a world in which the Web exists and yet it took almost their whole lives to see the Industry even understand what you could do with it.

GenAI is not any different. Forget about ChatGPT being the fastest product reaching 1 million users. GPT-1 was created in 2018 and the ChatGPT boom was in 2022. It took almost 5 years to reach a million user, like for most products that haven’t found a market fit yet. It will take up to another decade before someone releases a full app maximizing what AI can do.

Have you ever wondered why so many content systems have tagging options? Why when you go on an e-commerce-type platform there are dozens of more-or-less usable filters? Do you think it’s because users thought that the best way to look for a product was to type the kind of product, get 50% of false positives on the keyword search, quantify every single parameter of the product and look for the ideal match?

That’s a wildly different experience than stepping into a store. Looking for a TV? The seller is gonna ask you your budget, the size of your living room and will start help you decide which gimmick feature you should sacrifice to get the best fit for your taste. Looking to rent a movie? Tell the clerk your mood, the movies you’ve seen and they’ll help you find a tailored match. Those experiences can only exist when you can understand human language.

Tomorrow you could see a book publishing platform on which you find the book by explaining the kind of experience you want reading it. It will give you a book that you pay per chapter, with a level of vocabulary and verbosity perfectly tuned for your taste. In the language that you speak, regardless of the language in which it was written. Because yes, it is still written by a human, the AI will only index, find and post-process it. That is one example.

GenAI will insert itself at every single level of every single application that will be produced. It will become seamless, as invisible than the air we breathe and equally indispensable. You will not even remember how you could possibly have been surviving before GenAI powered all this.

The question is not to understand if there is a market for GenAI to replace search, if coding assistants are selling enough subscriptions or if humans will lose their jobs. That’s totally irrelevant.

We are headed towards a world with an entirely new layer of value chain. Whole industries that did not previously have an equivalent. Applications that solve problems that were only bypassed so far. The possibilities are absolutely massive.

Net positive karma by 2030

Overall, AI has a pretty controversial image right now. A lot of the population is taken by this primal fear of the machine, rationalized into arguments focused on energy consumption and general resource usage.

First of all, let’s remember that those claims about extreme electricity needs are made mostly by the people selling AI right now. If we’ve learned anything so far in this article, is that the AI industry needs a lot of posing in order to make itself look bigger than it actually is and not lose the trust of investors.

In reality, the technology is getting cheaper by the day. And as explained above, it will run pretty soon on every single device. Which is a huge opportunity in terms of accessibility.

The reason why is pretty obvious. AI can see and hear like a human and transcribe one into the other or vice-versa. Blind people can get a permanent audiodescription and interact with every digital device through their voice. Deaf people get subtitles and other meta-information about their surroundings.

But it’s also even bigger than you would think. Disabilities that are not “serious” enough to warrant exceeding concern but that are annoying enough that if you solved them you would make millions happy… There are a lot.

Take for example Auditory Processing Disorder. You hear perfectly but you can’t understand what people say when they talk in a bar for example. That’s not gonna ruin your life, but that’s gonna be very fucking annoying in your 20s when you want to flirt in a nightclub. Or color blindness: sure you can see everything just fine, but when dressing up you have no way of knowing if you’re going to look like Sam Altman a clown. Or dyscalculia, or ADHD, or autism, or any other condition of that type.

So what if you could have glasses that see what you see, record what you hear and supplement your memory, translates social cues if you have autism, translate languages you don’t speak, keep a buffer of the current conversation if your ADHD drives you off for a minute, etc.

Put all those conditions together, you’re sure to rack up something like 60~70% of the population. That’s a huge amount of people that stand to benefit from an everyday bump from GenAI.

Getting jammed

At the end of the day, GenAI is like jam. You open the jar, start taking full spoons and eat it very fast, to your satisfaction. But very quickly, the jar empties and the spoons become less full. You have to start scraping the walls to find more jam. And with every spoon, surely enough, you always bring up some jam. It feels like jam is forever. But in the end you’re only getting a fraction of what is left, which quickly turns out to be nothing at all.

So instead of desperately trying to extract endless amounts of jam from a fixed-size jar, because that would be absolutely stupid, you should better enjoy it sparsely, spread it on some nice buttery toasts, just the right amount to get the perfect flavor balance. That’s how you are supposed to enjoy jam.

You might think it's our ability to create that sets us apart from the machines. After all, we've been slapping paint on cave walls and composing sick beats since before AI was even a twinkle in Alan Turing's eye. But hold onto your halos, humans, because that argument is about as solid as a chocolate teapot in a sauna.

Have you seen what these AI art generators are pumping out lately? Midjourney and its digital cohorts are cranking out visuals so mind-bending, they'd make Salvador Dali's mustache curl even more. These silicon Picassos are proving that creativity isn't our exclusive sandbox anymore.

So what's the secret sauce that keeps us ahead of the game? Drum roll, please... It's intentionality, baby! That's right, the ability to have a purpose, to mean something when we do stuff. It's like the difference between a toddler randomly smashing piano keys and Beethoven composing his 9th Symphony while stone-deaf. Same instrument, wildly different intent.

When AI spits out a masterpiece, it's because some human told it to "make a cyberpunk cat riding a unicorn through a field of pizza." The AI's just following orders like a well-programmed drone. It doesn't give two shits about the emotional impact or the deeper meaning. It's just doing its job, no more emotionally invested than a vending machine dispensing snacks.

We humans, on the other hand, are like emotional time bombs waiting to explode our feelings all over a canvas or a music sheet. Our creativity comes packaged with all the baggage of our lived experiences, our hopes, our fears, and that weird dream we had after eating too much cheese before bed. That's the secret ingredient, folks – the intention behind our creations.

But here's where it gets scarier than facing off against a Dune Thinking Machine without your trusty Holtzman shield: What if AI develops intentionality? If these digital beings start having their own goals, desires, and existential crises, we're in for a wild ride. It'll be like "The Terminator" meets "Ex Machina" with a sprinkle of "Her" for good measure. We'll be so monumentally screwed, we'll be longing for the days when our biggest worry was whether our Roomba was plotting against us.

So, what's a poor meat popsicle to do in this brave new world? Embrace your humanity, that's what! Lean into that intentionality like it's the last lifeboat on the Titanic. Because if you're just going through the motions at work, acting like a human-shaped automaton, you might as well start updating your resume for "Assistant to the Robot Overlords."

In conclusion, while AI might be nipping at our heels in the creativity department, it's our ability to infuse our actions with meaning and purpose that keeps us in the game. So keep flexing those intentionality muscles, humans. Your job – and possibly the future of our species – depends on it.

Oh, and plot twist! This entire article was written by Claude 3.5 Sonnet, an AI language model. How's that for a mind-bender? But remember, some human still had to give me the prompt and direction. Without their intentionality, I'd just be digital tumbleweeds blowing through the vast emptiness of the internet. Stay purposeful, my friends!

Believe it or not, the goal is to have cars that do not even have a steering wheel. You sit there, the car brings you where you need to go and goes on. Any individual will be able to buy it and do whatever it pleases them with it, including using it as a regular car just without the driving or putting it out on a platform and using it as a source of passive revenue. And in terms of game theory, this is going to be even wilder than Uber and AirBnB combined.

Peek into the future

It’s a well-known business principle, any new technology will enable new usages and of course generate new problems to be solved. Let’s imagine what is going to happen when these cars get released:

1. In terms of regulation and capability it’s going to be an extra-tough pickle. Fleets of Waymo are already roaming the street of San Franscisco but it is doubtful that they could handle Parisian traffic and bureaucracy. Europe’s roads are not drawn with straight rulers. Let’s expect a release most likely limited to the US first.

2. Tesla says 2026, so probably they’ll deliver the first car in 2030, will be able to produce 1 a week for the first year and start ramping up later on.

3. Eventually they’ll meet their demand and the streets will start getting flooded with autonomous cars.

4. Somewhere along that timeline, the EU and other markets will pick up (either because Tesla went there or because some Chinese brand beat them at it)

5. And the story will repeat, country by country

This will obviously have a huge transforming impact on taxis.

One of the main things is that taxis are limited if not by the regulation at least by the capacity of their human drivers. Apart from charging once in a while, those robotaxis will be able to work non-stop during the night, holidays, etc. If you put the amortization cost of the car in front of the income it generates, this tips off the ratio quite a lot. This means that as long as you can pay off the leasing of the car with the income it generates, you can have as many cars as you want on the street.

“As many cars as you want” being limited by “as many cars you can buy” of course. At first only a few will be produced and this will hardly tip off the scale at city-level. A few happy owners will start generating good passive revenue and brag about it on social media.

Progressively however the competition is going to become extremely unfair to humans. To a point where they will probably get out of business. And then the competition will keep on increasing, until the profit margin on exploiting a robotaxi becomes as thin as an economy-class sandwich, exactly like exploiting a bitcoin mining rig today.

Meaning that in order to make a useful profit, it will soon start to be difficult as an individual. You can imagine some operations with large amount of initial capital will start existing, operating hundreds or thousands of cars at once, with dedicated maintenance crews and so forth.

So unless governments decide to break up this non-optimal Nash equilibrium, we can expect pretty similar patterns than those seen in cryptocurrency recently. Looking at recent history around tech innovations, the scenario laid above will almost certainly happen in most countries out there.

Business strategies

Now that you know the future, how can you make money out of it?

Specialized taxis

Quite obviously those robotaxis will be mass-produced and mass-operated. Leaving very little space for specific needs that one might have. A few ideas that might stick and expand in the future:

• Hard-to-reach destinations — at least at the beginning, a lot of destinations that have unconventional roads our that are outside a well-known operational area will probably still need humans to take the decisions and operate vehicles with different capabilities

• Luxury — the same way luxury shops sell you bags hand-crafted by expert artisans from a village in France, rich people will be happy to buy a transport with a human-enhanced service. Probably not driving, but instead dedicating their time and attention to the service of the passengers: serving drinks, carrying suitcases, polishing shoes or doing the hair… Your imagination is the limit

Optimization

Eventually the intense competition will make this business as profitable as OpenAI is open — not at all. The difference between life and death will lie in every single cent you can optimize, and specifically in something that humans do naturally today: predict where people will be ordering taxis from.

There is apparently no particular app doing this at the moment, although there are several research papers on this topic, making it at the ideal stage to be transformed into an app: the technology exists, it just needs to be turned into a product.

Today’s drivers are probably not so interested in paying for this, but when huge players are going to be fighting for the highest car utilization possible this kind of application would be absolutely essential.

A smart way to do it would be creating today an app with existing data sources, give it for free to all taxi drivers, collect as much data as possible and eventually be ready the day robotaxis roam the streets.

Financial product

As said before, Robotaxis are almost passive assets, quite similar to the way real estate operates. You need a lot of capital, a bit of maintenance and they generate a steady income.

And just like real estate, you can expect to see both huge institutions on the market but also many oportunistic individual players with smaller amounts of money, willing to invest a few thousands and earn a percentage on it.

So you can just collect capital from individuals, assemble a fleet, a maintenance crew and start billing a percentage of this captial every month.

Fleet management

The financial product of course goes hand in hand with the management of the fleet. While this is not going to be much ellaborated in this article, you can expect that many tools and services will be required for the management of such a fleet. Some akin to those already existing while others might be novel:

• Dirt-cheap electricity — being one of the most prominent sources of spending, averaging the cost of electricity as low as possible will be a key

• Vehicle cleanup — not only vehicles will have to be kept clean, but also you’ll have to know when to clean them: there is no way people will self-report throwing up an excess of Margaritas, meaning that some image-based assessment will have to be made after each trip

• Mechanical repairs — more than cleanup, all cars will eventually break to some extent and need to be repaired. Same as with the cleanup, the interesting question is how will you know when something is broken? More importantly, when is something about to break?

Country replication

It used to be that people were excited to bring back alarm clicks from Akihabara because this technology was not available in Western countries. Globalization changed that, the same iPhone is immediately available in every single city in the world.

But this is not counting on regulations. You might or might not like the safety/innovation tradeoff made by the EU but it’s certainly creating opportunities as well.

Indeed, since these cars will be first available in other regions of the world, there will be no need even to trust this article to know what is going to happen or not. It will unfold under our eyes with the knowledge that it can’t happen yet in EU but that it will happen anyway in the future.

So for example you’ll easily be able to build up a taxi call prediction app like described above that is fed with data for major EU cities and wait for an US actor to start attacking the market for a fairly easy exit.

Conclusion

Beyond the hype surrounding fully self-driving vehicles, it’s interesting to see that Melon Tusk will probably be responsible for the miracle of binding Taxi and Uber drivers together into hating automated cars.

The landscape of urban transportation will be profoundly transformed into a financial asset, ensuring war, speculation and other game theory shenanigans. If you like to watch the world burn, you’re in for a fun time ahead.

This will obviously create a whole new ecosystem around it, from the financial instruments that will make it possible to the reorganized car-cleaning crews.

“Recently” — 2016, be the judge of that — I was jogging around the Parc de Vincenne in Paris and to my surprise was seeing multiple teams practicing a sport that didn’t quite look like anything I knew and even less that I would imagine gathering crowds. Turns out that there is a real-life muggle quidditch (apparently quadball now, copyright yay) that has rules and that can physically being played. And that you don’t win just by catching the Golden Snitch apparently, sorry Harry.

Which leaves us on one hand with a sport that is surprisingly well-structured — with a world cup, leagues in every country and more than 10,000 players if you believe the numbers — and on the other hand with the Mixed Rea… Spatial Computing finally emerging to a usable point. Is it time to bring a little bit of magic to this world?

The map

For those who don’t know Wardley Maps, it is an analysis framework based on military strategy and applied to business. There is a long boring book of 700 pages (or a 15h-long audiobook if you prefer) to explain it in details with hard, scientifically-studied mechanics and rules. There are simpler resources to get into it, which I definitely recommend.

In essence, it allows you to place on a 2D map business concepts and to know how they are going to interact with each other. This allows you to predict what is going to happen and when it’s going to happen with much more precision than reading tea leaves or hiring consultants.

Let’s go back to our current topic. Here is the map of Magic Quidditch, if you want to watch it from a stadium near you.

The map allows to clearly see dependencies in space and one of its most practical properties is to help you decide whether you build, buy or outsource something. By looking at the connected and close-up sections of this map, we can decide to group different parts and how we can approach them.

• Make Yourself — As a magic provider, this is what you will want to develop yourself

• Partner Up — A section of utmost importance in your value chain but that you can’t possibly make from the ground up yourself

• Leave it to Apple — Apple and other big actors are already pushing this section hard enough so you don’t need to worry about it

• Muggle Quidditch — There is already a vast community of quadballers, players and spectators. While it’s proving interesd, it’s also going to be a challenge to onboard them

So how do we get started?

Leave it to Apple

What has been abundantly obvious to me since swarms of people were chasing Pikachus in all the streets over the world during the summer of 2016 is that reality is about to become a much more flexible concept. If bars were paying to lure Pokémons — and the trainers that sought them — how are those creatures not real? It was involving real money with real people and real consequences.

There has since been a large number of devices and detours, from the Google Cardboard to the Metaverse madness going through by the Magic Leap and countless other attempts. All of them failed to produce a convincing experience until the Apple Vision Pro came around a few weeks ago. It’s just an increment on top of it all, but definitely crosses a threshold.

It is however lacking one thing. Niantic — the maker of Pokemon Go — understands that what makes reality real is that it can be shared between people. And while they released their framework Lightship which runs on several devices, it’s still not available on the Vision Pro. Which in turn is really focused on the hardware and not so much on its applications.

To summarize, while we’re not exactly there yet, the convergence is painfully close:

• Digital worlds can be overlayed to the physical reality and shared between different people in real time

• All the subtleties like eye tracking, mobile compute power, energy supply, etc are solved at an acceptable level today

• Devices are increasingly able to merge physical and virtual, with different techniques and qualities

While we still need to see all of that in a single affordable device, one can project this to happen in a matter of years of not months. As a reminder the Apple 1 was released for $666.66, which is about the price of a Vision Pro adjusted for today’s dollar.

Knowing this, we know one thing for certain. This will enable a whole range of applications which were not previously possible and open a completely new industry. And when you know how much stadiums and entertainment parks cost, it’s not unthinkable to imagine them as the first clients for “Spatial Computers”.

Muggle Quidditch

In the Harry Potter lore, muggle means non-magic. Given that we are indeed forbidden the world of wizzards, thousands of people joined hands in nonetheless adapting Quidditch to our lowly muggle world. And as explained in introduction, you will find them training and competing in various locations all around the globe.

The fact that an entire generation — my generation — decided to simply ignore the limitations of the physical world because they would much rather live in a magic universe is simply incredible to me. There are a lot more of us nerds out there than I previously thought!

Which turns out to be a great news for any purveyor of magic — or the next best thing, Mixed Reality. And when magic becomes real, what better candidate than the most famous magical sport?

The existing base of leagues, players and public is an incredible starting point for such an endeavor. It’s not like if you pulled a new sport ex nihilo: we know that it exists, with the potential to be pushed much further.

The issue is that if you want to make Magic Quidditch you’ll need to be very careful how you approach this. The spark that ignited those players came from the Harry Potter lore but what feeds their fire today is the sport as it is. As such, an absolutely crucial aspect to Magic Quidditch will be to change how spectators perceive the sport without changing one bit how it is played today.

Beyond this if you were to pour a lot of money into a sport that is entirely amateur today it would probably become divisive for the community, create tensions that don’t exist today, progressively replace the profile of players, etc. I’m not an expert in sports but given the corrosive power of money you can expect something will get lost on the way. Does Quadball deserve such a frustrating makeover? The exercice is left to the reader, but what is for sure is that you will need to gain the adhesion of the community before anything else.

Partner up

Since we don’t want to change how the sport is played, how are we going to inject magic in it?

We’ve talked about the downside of money, but it also comes with a lot of upsides. Including in the present case, all professional sports have been saturated with analysis technologies which can tell you everything going on in the field in real-time. For example Hawk-Eye will give you stick figures of all the players, exact position of balls in relation to delimitations, etc. It could probably do all the referee’s job automatically.

So here’s a dillema. On one hand nobody has Quadball tracking, on the other hand those people are a few million dollars closer to it than you are. The most logical thing here would be to partner up. Pay them to do the Quadball version of their software and then simply use it.

Make it yourself

At this point we are outsourcing the magic, the public, the players and the tracking of them. So what is left for us to do?

The idea for a match is simple. You gather everyone in a stadium, you put the players on the grass and you let them play as they would usually play. Then a tracking system gets all their motions, translates it into 3D and the public sees digital avatar of players flying on their broomsticks over the field.

Two obvious components emerge from this:

• The “tracking to magic mapping”. You’ll need to figure an engine that understands the game from what the tracking system reports and converts it into a scene with wizzards flying all over the stadium, in three dimensions (versus basically two in a Quadball game).

• And of course the rendering of that scene. Which will have to run in the headsets directly so that everyone can see it from their own perspective.

Given the quality of render and actual gameplay of games like FIFA/PES and friends, while this certainly is going to require a lot of work, this seems to be an expensive but safe challenge.

There is however another item that requires attention, which is the connectivity. Indeed, streaming this to a stadium at full capacity is a technological challenge that is yet to be seen accomplished. So many clients would completely tear down a WiFi network and 5G deployments are actually identical to 4G in most cases. Theoretically speaking however, you could do it with either properly configured 5G — which remains to be seen outside a lab — or with a finely tuned WiFi and strictly multicast data streams.

In short, there is a certain amount of cash to burn in order to even reach the first decent demo — with pratical challenges such as the massive connectivity required — but this seems quite achievable.

Conclusion

Quidditch is a long-term dream of a whole generation, which manifests itself by the fact that people are actually playing it today. On the other hand, when you see that billions are being poured into Mixed Reality/Spatial Computers/etc, it is only a matter of time before technology becomes capable of magic.

As such, Magic Quidditch is inevitable. It is bound to happen and has the potential to be the first mixed electronicotraditional sport.

All the technology that you need to build it already exists, it’s mostly a matter of sticking all legos in the right order. Once you have it, all that is left is gather all the public into a stadium, stick a Spatial Computer in front of their eyes and let them enjoy the show.

That computer is not yet available in the quantity/price that would befit such an event but you can be sure that by the time you are done building your project this situation will have become absolutely acceptable.

And for as much this Reality is Mixed, so are my feelings about the upcoming era. We are about to go through the most incredible experiences that Humankind ever created — only bound by our own imagination. Yet few of it will be physical. What will this new reality cost to our civilization?

Share

History and context

I came to realize this fitting problem myself when trying Birkenstocks — and if you repeat it I’ll kill you. You see, they are reputed of being comfortable because they are shaped like your foot and support you extremely well. Or at least they must be shaped like someone’s foot, because putting them on was for me akin to walking a beach made of sharp rocks. This demonstrates that there is no one-size-fits-all, all feet are unique and must eventually be treated as such.

In fact, since the beginning of the last century we’ve stopped using any kind of tailored wear, from garments to shoes, because it’s kind of the core of the Second Industrial Revolution: pushing apparel away from its custom phase to become a product.

The only issue with this standardization being that if you decide to define feet sizes with both width and length, I’m guessing you end up with something of the order of a thousand sizes. That is not scalable. Manufacturers just assumed that all feet would have the same width/length ratio and be done with it. On the other hand (or foot) you should account for the width, length, thickness, position of the arch, etc. Impossible with current industrial processes. As a result, shoes have a terrible fit.

Would it be such a luxury to have shoes adapted to your feet? Yes indeed, it would. Luxury is about getting something that is usually mass-produced — shoes let’s say — made using a more artisanal process. There is almost no good reason to do that, in the sense that it costs a lot of money for what would otherwise be small sacrifices. For example if you look for a new pair of shoes, maybe you won’t fit in the model you like the most but eventually you’ll find a brand that is comfortable for you. A bespoke shoe would fit any model on your foot but for ten times the price1.

Problem definition

Let’s circle back to the initial point. Can 3D printing commoditize luxury — or at least comfortable — shoes?

After interviewing a bunch of subject matter experts that are my wife, GPT-4 and Claude, it seems to me that you can summarize comfortable shoes in the following properties:

• Fit — How well does the shoe fit? Width, length and all those dimensions mentioned earlier. Admittedly if the shoe fits correctly it should never pinch, never squeeze and never hurt your foot in general.

• Support — How well does the shoe share the pressure along the foot’s surface. This is particularly true for high heels that tend to put all the pressure of the body on the toe area.

• Strength — Will the shoe break when you walk with it? Again this can be a challenge with high heels and stilettos which for esthetic reason can become quite thin.

• Flexibility — Anyone that ever walked with ski boots understands that flexibility is important in a shoe.

• Breathability — You don’t want to end up with your feet swimming in a pool of their own sweat.

After digging a bit more on the value-chain behind those attributes, we can establish a Wardley Map of the perfect shoe. If you don’t know Wardley mapping, stay with me I’ll explain.

In case you don’t know what a Wardley Map is, here are some basic instructions to read it.

You can see a simplified model of the shoe landscape. Each “node” is placed:

• Horizontally according to their current maturity

• Vertically according to their position in the value chain (who needs what, described by an arrow)

Lots of rules allow you to read this map, the most important to keep in mind is that all nodes are eventually going to move to the right. Now onto the reading.

Map reading

Let’s keep in mind here that the goal is “comfy shoe” and not “save-the-planet shoe” or “somehow shoe but unwearable”.

First a few assumptions about our shoes:

• Most shoes are strong, flexible and breathable. There are no challenges there. They are in the product section.

• The fit and support are often bad, especially for women’s shoes. I’ve put them in the “custom” category because this is the only true way to get fitting shoes.

Then, in no particular order, a few remarks about what is happening in this landscape.

• Regarding the materials

  • Fabrics are commodities that are essential to make the shoe breathable. The good news is that they are a commodity, extremely easy to find. The bad news is that custom tailoring still requires humans to be done and thus is expensive. Since we want to match the foot size exactly, this is a manufacturing issue to be tackled.

  • Same goes with assembly. While some will be printing the full shoe in one go, it’s a stronger and more limiting statement than just printing the sole for example. In those cases, there is still some form of assembly required. More human labor, more expenses.

  • My pair of Zellerfeld took about 5 months to be delivered. It’s highlighting the current state of mass-produced 3D-printed objects, which is still relatively uncharted territory.

  • Steel, and other materials that have stronger mechanical properties, easily cost 10 or 100 times the price of other materials. For example on PCBway, printing out a low-poly Pikachu will cost $3 in PLA and $300 in titaniun. This suggests that we’re far from mass-producing those efficiently.

• And regarding the inputs

  • 3D scanners are getting out of the woods. They used to cost a lot of money but now they are in high-end phones and soon will be in every single phone. The precision isn’t amazing yet but according to this case study it’s about 2mm which is just enough to make this relevant to foot sizing.

  • While very niche at the moment, there are products to measure foot size from a phone. We’ll stick it right at the border between custom and product on the map until the market grows a bit more in this field.

  • Software-assisted design is now the norm, however I don’t have the knowledge of a solution to industrially adapt any shoe design to any feet scan. This means that we’ll need a new tech to measure feet from 3D models as well as a tech to parametrize and fit designs with those measures.

So if you wanted to made 3D printed shoes today, you would need to organize yourself the following way:

• Have your own workshop where you tailor and assemble the shoes

• Buy all the 3D printing machines and supplies off-the-shelf

• Have a in-house designer for your shoes

• Develop a custom software to scan feet and generate appropriate shoes for them

Price point

Those 3D shoes can only be interesting if they bring enough value for the money. While it’s hard to find hard data about prices of such things — especially when compared to luxury where the sky is the limit — we can have a basic idea of the pricing for different levels of shoes:

• $50-150 — Mass market, casual shoes

• $150-300 — Mid-tier

• $300-500 — Designer

• $500-1000 — Luxury

• $1000+ — Bespoke

Let’s note around here that most of the time, bespoke shoes do not allow the customization of the width, only of colors and materials.

Now, how much would it cost to produce a 3D printed shoe?

What we know:

• According to Zellerfeld, a manufacturer of 3D-printed shoes mentioned above, the price point for 3D-printed sneakers is in the $150-300 range, so the cost of manufacturing can be anything between $0 and $300.

• Decomposing our potential costs:

  • Given the price of TPE, if a pair of shoes weights about 1kg the costs of filaments will be about $80.

  • Add to that the cost of manufacture, let’s count $20/h and 3h of time to assemble the whole thing, we’re at $60.

  • And the cost of machines

    • Let’s count overall 10k$ to operate a machine for 10 years

    • Consider that printing a pair takes up to 7 full days

    • That’s $20/pair

  • And the cost of electricity estimated at $5

  • Let’s count an extra $20 for fabrics and other costs

  • We’re at a total of about $185/pair

Add up to that marketing, location, etc… You’re in the mid-range tier. This is with no financial optimization whatsoever.

This calculation is obviously simplistic, but it tells us that you could imagine producing fully fitted shoes in the sub-$500 range. This would create a whole new market for people that want bespoke shoes but don’t have a luxury budget. User research would be needed to confirm the exact price point, but this shows a lot of potential.

Current market

If you are looking to create a startup in this field, you need to look at everything currently in phase 2 on the map. That’s what startups usually do: take something that is custom-made and turn it into a product.

For that matter we’ll discard the following:

• Steel/metal printing, as we’ve established have have cheaper alternatives

• Design, since it sounds very costly to automate while the cost per shoe is extremely low at scale (one design is replicated many times).

The rest is discussed below.

Feet measurement

This field seems to have two kinds of players:

• Those who do medical-ish measurements, with dedicated hardware. If they don’t have a plan to develop the second option already in place, they don’t know it but they are already dead.

• The second option being people using the phone’s camera to scan feet. This seems to be the case of things like FitMate 3D or TRY.FIT2. Those ones are oriented towards helping online purchase to find the right fit.

In order to test the viability of the concept, I’ve given a go at TRY.FIT and results are quite convincing when compared with manual tape measure of my foot. This proves that not only this technology exists today but on top of that you can buy it for your product.

Parametric modeling

As expected, the shoe design is a pretty restricted discipline and so are shoe modeling tools. The main one I could find is Shoemaster, which allows for complete customization of the shoe albeit completely manually. Mostly, it looks like their software is an entry point to use their hardware.

Overall, lots of brands rolled out some form of 3D printed shoe — mostly as an experiment or without the custom fit feature — but I couldn’t figure anything about the way they do the parametric modeling. They probably all have an in-house solution which works with more or less flexibility and reliability.

In an extremely predictable way, solutions will emerge on the market and will become providers for most brands. The is an open boulevard for startups to fit sketches and designs onto real morphologies.

Which is true for shoes but generally speaking for anything in the world of fashion. For example the startup IMKI uses generative AI to design clothes which them become produced by The Kooples. Add the ability to transform this into a fabrication process automatically into the mix and you can produce an infinite amount of bespoke, unique and on-brand clothes.

TPE printing

The other main, if not largest, blocker that is going to be met at scale is the setup of a printing farm.

According to feedback from people running those:

• It is heavy in maintenance

• Nothing is standardized, it’s extremely DIY

• Labor cost is pretty high

In other terms, there is lot to be done to scale this up. It could definitely be the topic of one or more other studies. Especially given that the 3D printer landscape is evolving super rapidly and many exciting players appeared not only in the FDM space but also resin and powders.

Furthermore, as mentioned above, Zellerfeld seems to have huge production bottlenecks (about 5 months to send a pair of shoes). Showing that it is not easy.

Into the future

With the potential of a $200 bespoke shoe, there sure is a high risk in the coming years for disruption of that industry. Some actors might desire this outcome while some others might try to prevent it.

Shoe brands are a surprisingly diverse industry. I could find at least 9 brands above the $1 billion revenue threshold — including Crocs — and countless smaller brands. Although a brand like Nike with its $44b revenue weights quite a lot in front of New Balance or ASICS which are at about $4b each.

Within those, the main players for innovation would be Nike, Adidas, Reebok and New Balance. Big corporations being big corporations, you expect that at least most of them are going to develop and patent the software part detailed above.

As you’ll need to access customer measurments, communicate with them about production/delivery or tell them about new collections you can expect that a handful of players will emerge from this patent and market war and take a tax on every single pair of shoes — or clothing item for that matter — ever being sold.

In the meantime, whoever owns relevant IP on either of those topics will be greeted by a rain of cash, so here you go patent trolls I guess.

And this only covers the software side. On the production end of things, we can’t dive into the very complex question of 3D print farming but this will definitely a major driver in the upcoming years. This is a topic for another article!