Some people think of freedom as “doing whatever you want”. They want to speak freely and not have to filter their words, or want to have more free time and work less. They want freedom of choice but without the anxiety that comes from choosing, and perhaps also a free pass from the cost of a wrong decision. In other words, they want the upside of choice without bearing its consequences.

On an almost opposite side, I remember my high school philosophy professor defining freedom as “doing whatever you have to do, because you want”. His definition separates freedom from impulse and assimilates it to responsibility. It suggests that we are not necessarily less free because we have duties, obligations, or commitments, but quite the opposite we are free precisely if we have chosen those commitments so fully that they no longer feel imposed. Still, this definition can become dangerous if taken too far: a life built only around what one “has to do” can become noble to the world but exhausting to oneself, especially if responsibility is never examined against the life one actually wants to lead (see here for my thoughts about responsibility as a driver of life choices).

I prefer to define freedom as “the ability to make your own decisions and accept their consequences”. Both parts of this definition matter, and neither is straightforward. The first prerequisite for freedom is having the external possibility and enough inner capacity to make choices; the second is having the strength to live with what follows from them. A person who has formal options but cannot decide is not really free, and neither is a person who can choose but refuses the consequences.

Some limits to freedom are external and very real. If you are in prison, you are not free to move. If you live under a coercive political system, you may not be free to speak. If you are trapped by poverty, illness, violence, or overwhelming family circumstances, your choices are not the same as those of someone who faces none of these limitations.

For most people the constraints are fortunately less dramatic, but perhaps for that reason they are also easier to overlook. If someone lacks the strength of character to make choices, or if their happiness depends on pleasing others, their freedom is much narrower than it looks. Similarly, if every important decision is outsourced to their partner, parents, boss, friends, or social circle, or guided by whatever is fashionable that year, it may feel like they have plenty of options in theory but have none in practice. It is a strange form of captivity: living in an open-door cage from which one is unable to leave.

This is why the second part of the definition matters so much. One’s freedom is not only tested when making a decision, but also after it, when one path has been chosen and we need to accept the loss, discomfort, uncertainty, or responsibility that comes with having chosen. Choosing can be the attractive part; the harder part is living with the fact that every choice closes other possibilities. A person who wants to make decisions but not accept their consequences is not really free, nor is someone who decides what others expect because s/he cannot tolerate disapproval or because their happiness depends on pleasing them. For example, some people stay in a job they dislike because trying something else might expose them to rejection, lower status, lower income, or the uncomfortable discovery that they are not as prepared as they thought. Others remain in relationships they no longer enjoy because leaving would require financial independence, emotional strength, or the courage to disappoint people. Others keep living in a city, profession, lifestyle, or social circle that no longer fits them but remain there because changing would be too disruptive. All these situations have some variation of fear as a common denominator: fear of losing status or one’s standard of living, fear of being judged, fear of starting again, or fear of discovering that the problem was not only outside them. Moderate fear can protect us from making stupid decisions, when prudence becomes a permanent excuse for inaction, it becomes captivity.

Bar a few exceptional circumstances mentioned earlier, personal freedom is something we can and must build for ourselves without waiting for it to be bestowed by an external agent. For example, if I cannot leave a well-paid job because I have built a lifestyle that requires such salary, the company is not my only constraint: my lifestyle is also a constraint (remember this?). If I cannot speak honestly because my happiness depends on being liked by everyone, other people’s opinions are not my only constraint: my need for approval is also a constraint. If I cannot change the direction of my career because I have not developed useful skills, the job market is not my only constraint: my lack of preparation is also a constraint.

Our freedom is therefore built much earlier, through many decisions that do not look heroic at all, and which build our skills, health, savings, reputation, emotional independence, trusted relationships, and judgment. All of these qualities are inputs to our freedom, and create the conditions under which we can act when action becomes necessary: skills make us useful beyond one specific job, boss, company, or market moment; health matters because energy and resilience expand what we can realistically do; savings buy time and reduce dependence; reputation opens doors before we even need them; emotional independence matters because it gives us the courage to choose our wellbeing without collapsing; good relationships matter because they provide support without becoming a substitute for our own judgment; and judgment matters because few things reduce freedom faster than repeated bad decisions. They are all a form of stored freedom, and a sort of personal currency we can spend when facing complex decisions.

Most people want to be free, at least in theory, but wanting freedom and becoming capable of freedom are not the same thing. The first is an aspiration; the second is a discipline often built through boring decisions. It requires developing the strength to decide, the humility to accept consequences, and the honesty to recognize when circumstances are not the obstacle, but rather one’s own comfort or fear. None of it feels particularly epic while we are doing it, but over time it becomes the difference between leading a free life or not doing so.

What I found was a fascinating book and a deep analysis of human behavior and power dynamics, which could have been perfectly called “Treaty on human nature and how to gain and keep power”. It is Machiavelli’s manual on how to earn, conquer, and maintain a Principality, understood as a system in which one single person (the Prince) ultimately detains power. For context, Machiavelli had written about other forms of government, specifically Republics, in his Discourses on Livy – therefore The Prince should not be read as an advocacy for absolute-power regimes, and rather as an instructions manual on how to win when absolute power was at stake (a subtle but relevant difference). Because it was written in the 16^th century, The Prince must also be read in that century’s historical and moral context, and the power struggles that happened back then. But even if the moral principles that governed back then are far from what we consider today as acceptable, the book still has lots of learnings which could apply to (geo)political, business, or organizational contexts of the current world: replace the word “Prince” by “leader”, “Principality” by “organization”, “preserve” by “lead”, and a few others, and what you have is close to a leadership book written five hundred years ago.

As a fun fact, I was amused to also interpret the book as Machiavelli’s job application and motivation letter to work for Lorenzo de Medicis. From this point of, view, the book is also about what an intelligent, sidelined observer is willing to share in order to matter again. And maybe that’s as “machiavellian” as the content of the book itself.

Lots to digest, so let’s go part by part.

The Prince as a 16^th-century job application

Niccolò Machiavelli wrote The Prince in 1513, when he was 44 years old, under circumstances that were personally bleak and politically humiliating. Until just a year earlier, he had been a senior civil servant of the Florentine Republic. He negotiated with kings, popes, mercenary captains, and—crucially—observed power in action rather than in theory. However, everything collapsed for him in 1512. The Florentine Republic fell, and the Medici family returned to power. Machiavelli was dismissed from office, accused of conspiracy, imprisoned, tortured, and eventually released but banished from political life.

Machiavelli addressed the book to Lorenzo de Medici, ruler of Florence from 1513, and nephew of another Lorenzo from the same family: Lorenzo “the Magnificent” who had died in 1492. As the Medici then controlled Florence, Machiavelli knew that any return to relevance would have to pass through them. In this context, he was offering his knowledge and experience to help the new leader remain in power and rule more effectively. Although he is not openly begging for employment, the dedication of the book (“to bring Your Highness some token of my loyalty”) and its closing exhortation (titled “An appeal to conquer Italy and free it from foreign occupation”) are nothing short of an old version of today’s CV, a motivation letter, and a what-I-can-do-for-you proposal.

How to win and preserve a Principality (or an organization, in current terms)

The image above is a schematic summary of how Machiavelli classifies Principalities from the point of view of the new ruler and how he1 has acquired the new territory: whether by heritage, by conquest to integrate to an existing Principality, or by conquest to form a totally new Principality. For each of them, Machiavelli goes on to explain the behaviors and decisions that maximize the chances of keeping an intact Principality for long (which means the Prince holding power for long); these are also summarized in the schema. Again, think about how those schemas come to play in current-day organizations, for example how the leader of an organization, a business, or a country, acts when s/he comes to the top, and the decisions s/he makes to be successful and remain there for long. For a moment, make abstraction of the motivations of that person to take the leadership role: many of the resulting leadership behaviors will be agnostic of whether his/her motivations are noble (e.g., serve the country, make the whole organization successful) or ignoble (e.g., become rich at all costs, or feed one’s ego by remaining a leader for as long as possible, even of only for one’s own benefit and glory).

A modern equivalent of Ecclesiastical Principalities (apart from religious institutions themselves) could be organizations with a common and deeply ingrained set of principles and institutions. The more those values are shared, and the more widely accepted those institutions are to run the organization, the more stable they are when leadership changes occur. I would even go as far as saying that, in such organizations, the real power is held by the value system and the institutions, and the leader (or “Prince”) is the representative of those rather than their owner. One can think, for example, of centuries-old universities (e.g., Harvard), where the identity of the institution is far more durable than the tenure of any president or rector. Leadership changes occur, sometimes with noise, but the underlying academic norms, governance bodies, and traditions remain intact. The same applies to certain global institutions — central banks or multilateral organizations — whose legitimacy rests on codified norms, mandates, and cultures that constrain whoever happens to lead them. In such cases, the leader governs, but does not redefine the doctrine of the institution; s/he rather operates within it. And precisely because the doctrine is widely accepted, the system proves remarkably stable. In my professional experience, I have found that companies with the strongest (and best-shared) system of beliefs are not only the most successful, but also the most stable over time.

I assimilate Hereditary Principalities to family businesses, of which it is often said that “the first generation grows it; the second generation enjoys it; the third generation destroys it”2[2]. Statistics confirm that cliché: only around 30-40% of family businesses make it to the second generation, 10-15% to the third, and <5% to the fourth3. Machiavelli claims that to keep power in such cases, it’s important to not alter the order established by predecessors, and not to make decisions in a rushed manner. But this is not to be read as a call to inertia; rather as a warning against undermining the sources of legitimacy. In family firms, the leader’s legitimacy does not rest only on ownership, but on trust: the successor who behaves as if the company were a personal asset to reshape at will often accelerates decline; the one who understands him- or herself as a custodian of a legacy tends to preserve cohesion. The Egon Zehnder article quoted before alludes to it: “each succeeding generation sees the family business not as a matter of ownership, but of trusteeship”. For that reason, I see the most enduring family businesses are those that gradually move from a purely Hereditary logic to something closer to an Ecclesiastical one: shared values and deeply rooted institutions and governance. In those cases, the leader ceases to be a “Prince” by blood and becomes the representative of an institutionalized belief system.

Let’s now comment on Principalities that are new but will be annexed to an existing one. In old terms, this would mean a Prince conquering a new territory to incorporate it into his existing Principality. In modern corporate language we’d be essentially talking about mergers and acquisitions, and here the empirical record is strikingly aligned with Machiavelli’s messages, including that the hard(er) part is not acquiring but holding and winning long-term: research conducted over decades shows that in M&A acquirers tend to lose value around the announcement4, and some reports claim that 57% of acquirers ultimately destroy shareholder value5. On governance, the parallel is as direct: just like Machiavelli referred to eliminating previous ruling lineage when conquering adjacent territories, corporate takeover data shows that top executive turnover in target firms jumps post acquisitions (from roughly 10% annually before the deal to about 42% in the year following completion, according to this study). However, the evidence is a bit more complex than this simplistic purge logic: CEO retention can create value when managers possess integration-critical knowledge6, suggesting that in cases where the acquired company is not a “culturally close territory” for the acquiror, the better course of action is to retain credible local structures rather than do a full leadership overhaul. As Machiavelli suggests, when annexation fails it is not because conquest is impossible, but because consolidation is mismanaged; in organizational terms performance after an acquisition depends on aligning governance, cultural distance, and integration intensity.

Last, Machiavelli goes on to describe Principalities that are completely new i.e., those that are a result of a conquest and will not be annexed to a pre-existing territory. While I won’t go into every sub-case, I found a few ideas worth commenting. First, that taking and preserving leadership through one’s own capability is harder to do but easier to sustain, while when it is granted by others it is easier to obtain but fragile because if depends on their continued support. If you’ve ever been in organizations who rewarded promotion based on personal affinity rather than competence, you know perfectly what this means and how it impacts both the leader (whose authority is dependent on someone else) and the people under government (i.e., the teams who have to suffer the new leader). And second, that if you raised to a leadership position with the support and endorsement of the people you’ll “govern”, you’ll remain in power for as long you don’t oppress them, but it will be hard to sustain your position in the opposite case.

Power, morality, and “the end justifies the means”

As said earlier, one of the reasons I picked up The Prince was to understand the context in which Machiavelli allegedly defended the idea that the end justifies the means. The famous sentence never appears in that exact form, but the concept is surfaced in the book’s 18^th chapter as follows: “So if a leader does what it takes to win power and keep it, his methods will always be reckoned honorable and widely praised”7.

Is that morally troubling? Yes. But it is also descriptive rather than prescriptive i.e., Machiavelli describes how power works in rough environments: success often legitimizes methods whether we like it or not (have you ever heard the phrase “history is written by the victors”?). What the book forces us to confront is the uncomfortable idea that morality is unfortunately not self-enforcing: it survives because it is upheld by institutions, by shared norms, and ultimately by force. In the early 20^th century, Max Weber defined the modern state as the entity that holds the monopoly on the legitimate use of physical force. Following Weber’s definition, legitimacy and power are intertwined: values prevail not only because they are right, but because they are also backed by authority and (ultimately) sheer force.

For that reason, when the distribution of power changes, the moral vocabulary often changes with it. In recent times many have referred to a “new world order”, which might be nothing more (and nothing less!) than a shift in who has the authority to define what is acceptable. A recent interview to current U.S. President Donald Trump captured this starkly: asked whether there were any limits on his global powers, he replied: “Yeah, there is one thing. My own morality. My own mind. It’s the only thing that can stop me”8. The statement is striking not because it is novel, but because it is explicitly Machiavellian.

Five hundred years later, The Prince remains relevant not because it defends or celebrates ruthlessness, but because it is a blunt reminder of how humans may operate when it comes to reaching and preserving power. From that perspective, Machiavelli does not tell us what should be; he tells us what tends to happen.

This cycle describes my last… (ehem) many years surprisingly well. And judging by the nodding heads whenever I’ve shared it with others, I suspect it’s a wheel many more people recognize themselves in. I don’t want to say they are trapped in, because in most cases this wheel is the result of personal choices: we are not necessarily trapped in any one single stage; rather, we go in circles around it, and trying not to fall sick while spinning.

I won’t pretend I have a magic solution for how to steer or stop this wheel – not for myself, and therefore even less so for you reading this article. Tons of ink has been poured already on work-life balance and work-life harmony (see from 1:30 if you’re unfamiliar with this concept). Yes, I’m aware that virtue stands in the middle, but that is easier said than lived: “working hard” and “enjoying life” seem to drift apart in the busy-ness of everyday life, and aiming for a theoretical middle point may deliver neither the enjoyment nor the impact you’re after. I’m also skeptical of claims of the type “I love my work, therefore I’m already enjoying life”: you can absolutely love your work, but that does not magically turn 24-hour days into 48-hour ones so that you can also spend abundant time with family, exercise five days a week, read widely, nurture friendships, travel, and master that semi-abandoned hobby of yours – all of which I associate with enjoying life. I also won’t moralize about whether this is a healthy cycle: it almost certainly isn’t, and I wish I didn’t think about it as much as I do. But it exists, and rather than ignore it, the best we can do is try to understand it.

What I’m most interested in is the why’s of the cycle: why the wheel keeps on turning and what are the mindsets that push us from one position to the next. I believe it all starts from one’s life goals (implicit or explicit) and from there boil down to a small set of fundamental motivation drivers: enjoyment, wealth, transcendence, and responsibility.

Let’s tackle them one at a time.

Life goals

Our lives are governed by a strikingly large number of goals. Some might be explicitly stated, even publicly communicated; others are so implicit that we may not even realize they exist. Yet all of them shape how we allocate our most finite resource: time.

Work-related goals tend to be visible and explicit: reach X in sales, deliver Y in profit, raise customer satisfaction to Z, launch N projects. Work goals are often defined annually and reviewed quarterly or monthly. Personal goals, by contrast, tend to remain implicit. We all have them, but rarely write them down or even notice them unless we fail badly in reaching them. They range from the most mundane stuff (get the kids to school on time, pay the mortgage) to the aspirational kind (exercise more, lose weight, read more).

However, rarely do personal goals have a time horizon greater than a year. If I asked about what goals people have for their lives (not for the next year), I suspect many would hesitate or improvise something broad on the spot e.g., “be happy”, “be healthy”, or “have a loving family”. The irony is that without conscious effort, our lives end up governed by a series of goals with very short lifespans and which we might even never have articulated. We want to be in control of our lives when we are not necessarily stating or managing our life goals, and that mismatch breeds frustration.

Instead, it is much better to define your life goals upfront, and do so crisply manner. For some, their goal is to enjoy life. For others it might be to accumulate wealth. For some it might be to leave a legacy (transcendence). And some may walk life “doing the right thing”.

Whatever your case, your goals (explicit or implicit) should be a reminder of the direction in which you want to lead your life, and support you in finding your right spot in the work harder vs. enjoy life cycle.

Motivation driver #1: Enjoyment

In philosophy, Hedonism (from the Greek word ἡδονή, read hēdonḗ, meaning pleasure) supports “the belief that the most important thing in life is to enjoy yourself”1. Hedonists place pleasure at the center; for them, pleasure is the highest good and goal in life, and pain is the ultimate evil. In its simplest form, a good life is one that maximizes pleasure and minimizes suffering. Epicureans, more restrained, argue that the highest pleasure is not intense enjoyment, but the absence of pain and mental disturbance (ataraxy). Same word — pleasure — and radically different interpretations.

But this isn’t a philosophy lecture: the point is simply that the role of enjoyment or pleasure as a life goal is an old question, unresolved, and deeply personal. There’s not a right or wrong answer: you should go for whatever most resonates with you. In my case, there are two opposing forces: one pushes me away from enjoying life, and the other back again into it.

Pushing me away from enjoying life is the tension between enjoyment today and enjoyment tomorrow. You may recall the fable of The Ant and the Grasshopper from your childhood. Or maybe you have heard about the marshmallow experiment run by (at the time) Stanford student Walter Mischel in 1970. In this experiment, children were handed a marshmallow and offered the choice to either eat it (and they’d only have that one treat), or wait for 15 min until the adult returned and would give them a second marshmallow. Many chose the smaller immediate reward instead of the larger delayed one. In later studies, the researchers found that children who were able to defer gratification i.e., wait and get a greater reward, tended to do better in life. I have a tendency to defer gratification (or “enjoyment today”) in favor of a higher reward tomorrow. However, such mentality comes as a cost. No less than 30 years ago, a high school teacher reminded us that “living in the past causes depression; living in the future causes anxiety”. The sentence did not offer a so what, but one would conclude that his recommendation was to “live in the present” (i.e., Carpe Diem), prioritizing enjoyment in present time through moments of pleasure like laughter with our family, kids, or friends, so that by the time we’re old our life journey will have been worth it.

Pushing me back into enjoying life is the realization that good relationships keep us happier and healthier. That’s actually the conclusion from the Harvard Study of Adult Development, which I already mentioned here and for which you can watch a summary in this video. As I already mentioned back then, relationships take time and effort, which means that, unless you invest in them today you won’t enjoy them tomorrow. It will be harder to nurture those long-term personal relationships (which will usually happen outside of work) if you’re permanently in a work hard mode.

Motivation driver #2: Wealth

Wealth as a motivator is not necessarily about being rich: in this context it’s about having one’s needs and desires covered, at whatever level one defines as enough. For some, that means paying debts and achieving basic financial stability. For others, a comfortable lifestyle e.g., a nice house and car, good education for your children, or a fair dose of leisure. For others still, maximizing income to preserve optionality. For a few, to accumulate assets without a clear endpoint.

In all cases, this is closest to “working for money”.

I tend to see this as a constraining motivator, which adds obligations and weight onto our shoulders. It will generally push you to the work harder zone of the cycle, and in an extreme case put you into a fast-spinning rat race. Beyond a certain threshold, money rarely liberates. It mostly locks you into systems you now have to sustain (see my earlier post, “When assets become liabilities”).

So the question is not whether wealth and financial security matter — they do — but how central they are to your life goals, and when “enough” yields to “more”, to the point that it makes you lose your balance between working hard and enjoying life.

Motivation driver #3: Transcendence

Transcending is defined as rising above or going beyond the limits, but in the context of this post, I prefer to define it as having an impact in the world, which outlives you. As a life goal, I associate transcendence to achieving something that is bigger than you, and the impact of which lasts after you’re gone. It is not about reaching fame: cemeteries are filled with people who were famous in life but were also soon forgotten after their death. Unlike wealth, I see transcendence as a positive motivator: although it also adds weight to your shoulders, the sense of purpose makes the weight more bearable.

In the context of this post, the question is not whether you want to transcend as a life goal, but how: if by doing, or by being.

You can transcend by doing, building something meaningful that outlives you e.g., a work of art, a book, music, or a sustainable business. Transcending by doing pushes you towards the work harder zone, as you’ll need to perfect your skills and overcome whatever difficulties lie between you and that transcendent work.

But you can also transcend by being: creating a lasting, emotional, intangible impact on others e.g., learnings, memories, lessons, or moments that others close to you will cherish when you’re gone. If you’ve seen Disney’s movie “Coco”, you know exactly what I mean. Although that impact builds on all the experiences you have accumulated over time (including the ones from work), transcending by being does not push you (as much) towards the work harder zone of the cycle, as those lasting memories tend to live among your close personal relationships, rather than work-related ones.

Motivation driver #4: Responsibility

Responsibility is about doing what needs to be done, not because it is enjoyable, profitable, or meaningful, but because it’s required or expected from you. Family duties, professional responsibilities, social expectations… all different flavors of this same concept.

Responsibility is a noble principle: it comes with commitment, reliability, and care for others. And precisely because of that it can become a particularly sticky motivator: you can morally question the pursuit of pleasure or wealth, and can lose faith in transcendence, but it’s hard to challenge the goodness of being responsible.

However, it can also become a dominant life motivator almost by stealth. And when it does, it often makes you act based on what you “owe” (to people or to systems you are part of) rather than based on what you “want”. Without a deeper purpose, responsibility pushes you towards the work harder part of the cycle, its weight accumulates, and (in a worst case) leads to burnout. The question, therefore, is not whether responsibility should guide your life, but whose expectations you are being responsible to, and whether they are aligned with the life you want to lead.

Concluding

The motivation drivers above are not mutually exclusive. They may also change over time – I expect many people are motivated by enjoyment when young, wealth in their middle age, and transcendence in their later part of their professional life.

Coming back to the wheel I started with, it is probably the tension and the shifting balance among those forces that keeps us spinning between the work harder and the enjoy life positions. And maybe that’s the point. Perhaps the paradox is that, while many of us want to stop or balance the wheel, it only works when it is moving, like satellites in orbit: slow them down and they fall; speed them up too much and they escape into space. Have them spin at the right speed, and they’ll remain where they’re meant to be, fulfilling their mission.

In studying land ownership in Italy, he observed that roughly 20% of the population held 80% of the wealth. The pattern kept appearing in other domains, and Pareto’s name eventually became attached to what we now call the Pareto Principle or the 80–20 rule: the idea that a small proportion of causes often drives a large proportion of the effects. The Pareto distribution is one of the earliest-known power laws i.e., statistical distributions where small events are extremely common while large (but rare) events account for most of the impact. In the picture below which illustrates a typical Pareto curve, a few events (numbers 0 to 20 in the horizontal axis) generate most of the outcome (80% in the vertical axis), while the remaining long tail of events (numbers greater than 20 in the horizontal axis) struggle to lift the Y-axis curve further up. The first part of the curve reflects highly productive events (few, but with high impact), while the latter part reflects diminishing returns (lots of events, little impact).

This pattern does not only manifest itself in wealth distribution, but also appears in multiple other orders of life. If you own a business, a small portion of your customers are likely to deliver most of your revenue. In production, a small number of steps are likely behind most of your output errors. Even in your current daily life, it is likely that a small number of apps on your phone make up for most of your phone time.

From here, it is tempting to conclude that we can often obtain 80% of the results with 20% of the effort and therefore we should focus mostly (or only) on the 20% of activities that deliver most of the value. But this is a misunderstanding: Pareto never proposed the 80-20 rule as a management technique or a recommendation to optimize resources. He was describing a recurring pattern, and not prescribing a way of working.

If you run a production line and 20% of your machines consistently cause 80% of your production failures, it makes sense that you try to fix the problematic 20% first; in this case your priorities match the Pareto distribution. However, if you’re an investor and find out that 20% of your stocks are delivering 80% of your returns, it may not be a good idea for you to sell the large chunk of underperforming stocks and double-down on the successful 20%. Unless you’re totally certain that the 20% will continue to outperform the market (which no one ever is), you’d be unnecessarily increasing your investment risk profile. In this case, Pareto’s 80-20 describes a backward-looking distribution, but cannot be taken as a forecasting tool. Likewise, if you’re a restaurant owner and 20% of the dishes in your menu generate 80% of your revenue, it would be a bad idea that you eliminate the less popular 80% of the menu. You don’t need to have infinite options, but even less successful dishes are still important to attract new customers, cater for diverse tastes, or just encourage repeat visits.

Again, in all these cases Pareto’s rule describes the distribution of results but should not be seen as a recommendation for prioritization, and even less a forecasting tool.

The consequences of erroneously using an 80-20 approach as a prioritization principle compounds in multi-step systems. When multiple teams intervene sequentially across a value chain to deliver an output, having each of them focus on the most productive 20% may be catastrophic.

For example, consider a project that requires the involvement of just three teams: Product, Design, and Engineering. The Product team imagines their ideal product, but knows that going after every functionality won’t fit into their target sales price. They therefore decide to focus on the 20% of functionalities that deliver 80% of the expected value. Now it’s the turn of the Design team: they love the product idea, but even in its simpler concept there are physical constraints (shape, weight, aesthetics) that make it impossible to design. They therefore decide to focus on the 20% key features that will deliver most of the product specs. Last comes the Engineering team: again they see the point of what the Design team has provided, but it is impossible to manufacture: they can reasonably deliver 80% of the features, but need to drop the most complex 20%.

By the end of this process, the 80% proposed by the Product Team has become 64% after Design (80% x 80%), and 51% after Engineering (64% x 80%). Each team prioritized correctly (or apparently so), but the customer receives barely half of the initially desired features. If instead of “only” 3 steps there were a hypothetical 7 the loss would increase to almost 80% (i.e., only ~20% gets delivered); if there are 10 steps the loss of value is almost 90% (i.e., only 10% gets deliver). What looked like a logical prioritization becomes self-sabotage just by compounding successive 80-20’s.

Worse still, the effort required to recover the missing portion is enormous. To move from the 51% to the full 100%, the organization must deliver all the dropped functionalities at each stage across Product, Design, Engineering, and beyond. And let’s recall that although they “only” accounted for 20% of the benefit, they represented 80% of the effort at each step, which is a huge gap to overcome. Each team had good reasons for deferring that work, but the collective impact is a massive gap that cannot be patched cheaply.

The issue with Pareto as a prioritization tool is not that the principle is wrong, but that our use of it can be lazy. You cannot optimize in isolation when you are part of a larger value chain; if everyone did the same it would just trim outputs to an unacceptable level.

Instead, an 80-20 prioritization in large(r) organizations looks more like a well thought-through intersection of each team’s work, where the most productive collective output is achieved by overlapping each team’s focused efforts. If you are familiar with the concept of “Ikigai”, prioritizing in large organizations is metaphorically closer to finding their Ikigai (if you’re not familiar with the concept, I strongly suggest you take a few minutes to read about it here or here). However, this requires a very conscious effort which can be highly complex and time-consuming when multiple teams are involved; I actually discussed here about how complexity in organizations lies precisely in the interaction among teams. In moments of organizational urgency or scarcity, this may feel like an unnecessary and expensive step. But (believe me): it won’t be as expensive as a lazy misuse of Pareto’s 80-20 rule.

1. You own your career. Not your manager. Not HR. Not the company you’re in. A great manager can empower you, open doors, or advocate for you, and that’s invaluable. But your manager does not own your career. Nor does any organization. You need to be the one defining what career you want, charting your course, and overcoming whatever hurdles appear along the way. You’re the one who must ask for feedback, make decisions, take risks, and sometimes say no. Waiting for others to develop your career is a fast way to stall it.

2. Accept that a career is made of ups and downs: it is never a straight line. Careers have plateaus, detours, moments of fast acceleration, and sometimes hard stops –even crashes– but they are never linear. Even high-profile careers are full of stumbles. Steve Jobs was famously fired from Apple before returning years later to reinvent it. While that’s a dramatic case, most careers have less visible setbacks: a reorg, a bad fit, burnout, or personal crises. Accept the highs and the lows: the key is to stay in motion. Learn from mistakes but don’t overinterpret them.

3. A career is a long-term “project”, not a short-term proxy. I already wrote about it in an earlier post. When thinking of your career it is easy to focus on your next job or job title. Instead, you should ask yourself what kind of professional you are becoming, whether you’re making progress towards your goals, or what you want to be known for X years from now. In other words, whether you are putting the right career inputs to maximize the chances of achieving your desired career outputs. Outputs are what get the attention, but the most important part of career-making often goes unnoticed, happens beneath the surface, and requires patience and perspective: it happens when you work on inputs.

4. Maximize your professional value, not just your salary. Chasing compensation is tempting, and in very specific situations can even become necessary. But the real asset you’re building is your professional value: your capabilities, your reputation, your network. Those are what sustain long-term success. I expanded on this idea here.

5. Strive to make yourself redundant. This sounds paradoxical, but I’m convinced of it. First, recognize that market forces will push for your redundancy whether you like it or not: new technological developments, competitors, or a company’s own drive for efficiency. You have a much better chance of thriving in the long run if you’re the first one to challenge what you do, how you do it, and how you add value. And second, it is false job security to base your career in making yourself indispensable. If the place can’t run without you, sooner or later you’ll become a bottleneck and struggle to scale, and progress. Metaphorically speaking, you’ll be trapped in your castle, unable to cross your own moat. Instead, build systems: document, delegate, and mentor others while you also think about what’s next for you. The more you empower others, the more space you create for yourself to grow.

6. Stay curious. Industries shift, tools evolve, and skills decay – in that context, curiosity is your career’s best insurance policy. Stay alert to what’s changing around you and what lies ahead of you. Ask yourself how your strengths are aging, and what blindspots you may have. Remember that what got you where you are won’t get you to where you want to be, especially in a world where technological developments seem to deliver a month’s worth of change every week.

7. Accept ambiguity. In your career you’re playing probabilities, not certainties. You can do all the right things and still not get the outcome you hoped for. The project you’re working on might be shelved, a reorganization might hit, or your manager might leave. Even if luck plays against you, that doesn’t mean the effort was wasted: control what you can and get comfortable with what you can’t.

8. Fit matters as much as skill. You can be an excellent professional and still underperform when put in the wrong context. Conversely, you might thrive somewhere not because you’re the best, but because the environment fits you best. When choosing where to develop your career, don’t just chase prestige or roles but also pay attention to culture, values, or how the role matches your strengths. Sometimes success is about belonging.

9. Career success is a combination of talent, effort, and luck - in unknown proportions. You need talent to stand out in your career (it’s hard to succeed without it) but talent alone isn’t enough. You also need to put in the work to feed and develop your talents – otherwise others will outwork you and compensate lack of talent with effort. But even then, outcomes also depend on luck, and luck can take very different forms: timing, sponsors, projects, or economics, to quote a few. As mentioned earlier, you can’t control luck, but stay humble, work hard, and keep showing up. You never know when the stars might align.

10. Your career matters, but your life matters more. If you’ve gotten this far in my post, you probably care about your career; I certainly do. But your career is just one part of your life. Make enough room for the other parts: health, relationships, your own mental well-being, or whatever you value in life. Remember that when life gets hard (and it will, sooner or later) the strongest support you’ll get won’t come from those around you at work, but rather from those around you in everyday life. If your identity is entirely tied to your job, that should be a red flag. As I sometimes put it, I don’t want my epitaph to be “He made it to [whatever title] at [whatever company]”.

For a moment, picture yourself living in an apartment building. You bought your flat a few months ago — maybe even 1 or 2 years already. Mortgage, boxes, unpacking… the whole thing. You certainly plan to live there for a while — perhaps even raise a family. As with any apartment building, you have neighbors next door, upstairs, and downstairs. They’re not necessarily your friends — probably never will be — but you see them from time to time and have a polite relationship: “good morning”, “good afternoon”, “looks like it’s going to rain, huh?”, and so on. One day, not long after you’ve settled, an upset neighbor knocks on your door, complaining about the noise coming from your apartment: you like to organize late dinner parties, and you're still doing some renovations on weekends. How would you approach that conversation?

Now, let’s change the context. You still live in the same apartment, but this time you’re about to move out — in a couple of months, you're gone. The same neighbor knocks on your door with the same complaint. Would you approach the conversation any differently? And do you think your neighbor (who knows you’re leaving) would do so as well?

Although a lot of factors may come into play, it's likely your willingness to compromise and find a mutual solution will be greater in the first scenario than in the second. Knowing you’ll soon move out and never have to deal with that neighbor again makes it easier to ignore their complaint and carry on with your noisy lifestyle. Similarly, your neighbor might become less inclined to engage constructively. Depending on their personality, they might go irate (“I’ve been putting up with your noise for too long!”) or passive (“Whatever, you’re leaving soon anyway”), but not necessarily collaborative.

The same odd dynamics often show up in business. Imagine it’s not two neighbors, but two companies in a business negotiation. In the first scenario, both sides know the relationship is long-term. After this negotiation there will be another, and another — it’s an ongoing, indefinite engagement. How willing are they to approach the next negotiation in a collaborative, win-win mindset, rather than trying to extract short-term advantage at the other’s expense?

Now imagine both parties know this is their last negotiation together. The outcome will have no future impact on their relationship. Would they act differently, and even be willing to betray one another to capture more value? If they believe karma can’t hit them back — that there are no reputational or practical consequences — the temptation to act selfishly and “win” at the other’s expense increases.

A British TV game show called Golden Balls (aired between 2007 and 2009) gives us some wonderful illustrations of this. In the final round, two contestants compete for a cash prize in a zero-sum game. Each is given two balls — one labeled "Split", the other "Steal" — and must secretly choose just one to play. If both choose Split, they share the money 50/50. If one chooses Steal and the other Split, the stealer gets everything and the splitter gets nothing. If both choose Steal, both walk away with nothing. Before revealing their choices, players have around 30–60 seconds to talk, plead, promise, negotiate, or bluff — trying to convince the other to split while secretly deciding whether to split or steal themselves. It’s a version of the Prisoner’s Dilemma, but with real money, real pressure, and often real betrayals. If you're curious, you can watch a few of those moments here, here, or — my favorite — here.

Let’s push this idea one step further. Imagine you're negotiating with a business partner — client or supplier — and you know (or suspect) that they might take advantage of you in your last deal together. But this isn’t the last one; it’s the second-to-last. Would you consider taking advantage of your counterpart already this time, knowing that you expect he’ll take advantage of you in the next one? And if that’s the case, what might be your approach not just for the second-to-last negotiation, but for the one before? Just imagining an end date to a relationship can shift how both parties behave, and how much effort they put into collaboration vs. self-gain.

At risk of stretching the argument, let’s look at marriage through the same lens. Long gone are the days when marriage was legally meant to be a lifelong contract. And regardless of one’s religious beliefs, I think it’s a good thing that people can leave a relationship that makes them unhappy. But… is it possible that knowing divorce is a (relatively easy) way out of a relationship, that option makes some people invest less in building a successful marriage? I remember a few years ago someone — half-serious, half-joking — suggested that marriage should work like a fixed-term contract, to be explicitly renewed every few years. Leaving one’s moral convictions aside, I wonder if the need to “negotiate” a marriage renewal with your partner every few years would increase or decrease the number of separations. Would such a setup make people try harder, or give up sooner?

I do not intend to use this post to offer lessons on negotiations or marriage. And let me be clear, I’m not endorsing any behavior by which it is OK to betray the trust of your counterpart: neither in your last deal, the second-to-last one, or any before that. Quite the opposite, my intent is to underline that in a situation where incentives strongly push you to maximize self-benefit to the detriment of someone else, it’s better to resist the temptation of abusing the other party and instead you should collaborate to the same degree as you would in an ongoing, indefinite relationship. Because in life, there are almost always downstream consequences.

“Karma is a bitch”, they say. And it has no deadline.

I first heard about this story and the expression “Poincaré Point” in this post by Jaime Gómez-Obregón, and read again about it on this other post by David Bonilla (full text available here, in Spanish). Both reflected on how computing has gone beyond the point where it is possible for one single person to comprehend the whole discipline, and how those born in the decades of 1970 - 1980 belong to that frontier generation.

In addition to math and computing, we have probably surpassed the Poincaré Point in many other fields like medicine, history, or physics. This might feel nostalgic for those belonging to a frontier generation, frustrating for those coming afterwards and eager to know it all in a particular field, and exciting for those who accept it as a as a sign that our collective knowledge increases, and with it our ability to build a better future.

But irrespective of how it makes you feel, is there a risk when a discipline or an organization surpass the Poincaré Point? I believe so.

This image (source) illustrates that risk metaphorically, and yet perfectly. It is the risk inherent to a discipline which grows, and in growing develops more and more ramifications and turns complex; so complex that at a certain moment it is impossible to be grasped by one single person. Still, it continues growing in new and edgier directions, more attractive, more stimulating, more lucrative. Because no one is capable to comprehend the discipline in its entirety anymore, and because most resources are channeled towards developing the edgier parts of it, it may occur that no one looks at parts of its core, nor is able to understand them anymore. At that moment, a blind spot is generated. An area that no one knows how to maintain anymore, that no one remembers how it works anymore, or (even worse) that no one knows it exists. The blind spot makes the whole set fragile and vulnerable. And that’s where the risk is.

Obviously, not all disciplines are subject to such risk in the same way. It is unlikely that a math theorem will be forgotten, or that it is no longer studied or understood. But not being an expert in programming, I was shocked when a few years ago I learned that COBOL, a programming language born in 1959, was still behind core systems in many big actors of the worldwide economy, and that there were few professionals capable of managing the systems. Worse, replacing COBOL systems was a huge mess. Maybe that is not (yet) a blind spot, but it certainly is a fragile spot.

Some will claim that “one’s risk is someone else’s opportunity”. They’ll claim that blind or fragile spots are not a problem because, if there’s a real demand for such fundamental knowledge (e.g. COBOL), market forces will take care of the issue and create supply for such professionals. Unfortunately, I’m not that optimistic: markets are imperfect and tend to anticipate scenarios that are possible and likely. They are less efficient when it comes to anticipating possible but unlikely events. But they’re not good when it comes to anticipating the unknown. And because blind spots are areas no one is aware of, by definition they fall into the bucket of the unknown.

The risk that occurs at industry level can also happen within an organization or business. While simple in their early days, companies tend to become more complex over time, especially as they grow in people, breadth of competencies, physical resources, systems, and processes. Early employees might still be able to keep a comprehensive view of the company, but newer employees will only develop a narrower view of their individual area. In parallel, companies continue exploring new business opportunities to feed and sustain growth: resources are channeled to fresh new areas with an appealing business case, potentially to the detriment of maintaining or upgrading core but apparently less productive areas.

Over time, people move, competencies get lost, and systems and processes change. One day some of those competencies, systems, or processes become a blind spot about which there is limited knowledge (like the earlier example on COBOL), or even worse no awareness at all.

Especially in large organizations, the vast amount of collective and distributed knowledge is both a blessing and a curse. It is certainly what allows the company to deliver sophisticated products and systems. But it also means very few people (if anyone at all) will have understanding of potentially core systems in full, end to end. On top, interdependencies among different areas of a company may mean that responsibility is spread or diluted. You lose the notion of what you know and what you ignore, and it’s easier for those blind spots to appear. As mentioned in this recent post by software expert Daniel Lemire: “When systems grow so intricate that no one fully grasps the foundations, we risk becoming like children inheriting a world we do not understand—wherea things work, or do not, and no one knows why”.

In summary: surpassing a Poincaré Point in a discipline or in an organization is unavoidable and even positive to the extent it is a natural consequence of its growth and development, but it is not exempt of risk because of the surge of blind spots.

While strong disciplines, industries, or organizations will be able to reduce that risk, capturing and codifying much of the collective experience as institutionalized knowledge, weaker ones will see core areas of expertise rely on tribal knowledge that can get lost over time. The greater the complexity, the greater the risk, and the greater the effort that the organization needs to put in maintaining a robust core.

If my gesture was comparable to other addictions, by what mechanism could that gesture develop and what could be downstream consequences?

1. Dopamine as a starting point

Dopamine is a chemical in the brain that plays a key role in how we feel pleasure, motivation, and reward. Think of it as a little "pat on the back" that your brain gives you when you do something it perceives as rewarding or beneficial. It helps reinforce behaviors that are necessary for survival: eating delicious food, having sex, or achieving a goal are all activities that release dopamine. Our brains also release dopamine when we connect to each other, which incentivizes us to do it again and reinforces our identity as social beings. However, dopamine can also play a role in impulsive or addictive behaviors when people chase dopamine in unhealthy ways like smoking, taking drugs, or addictive gambling. Dopamine itself is not the cause of addiction, but rather the reinforcement mechanism that keeps people coming back for more.

Dopamine-generating stimuli are widely accessible as a result of affluence and overconsumption in today’s society. That is at least a core argument in Stanford psychiatrist Anna Lembke‘s book Dopamine Nation: Finding Balance in the Age of Indulgence1. Because of such abundance, our brains respond by actually decreasing dopamine transmission, which creates a chronic dopamine-deficit state, and in turn makes us crave for even more dopamine. In simplified terms, overconsumption of dopamine-generating stimuli has made us addicted to such stimuli. Stating the obvious, dopamine is not bad in itself, but quite the opposite. It is uncontrolled dependency on it which is a problem. In my personal anecdote above, the instant gratification that social media provided (feeling connected to others, or consuming fast bits of engaging content), is what got me hooked to it and made me compulsively crave for more.

2. Technology enables scale

For decades, and especially during economic expansion post-WWII a.k.a. “the golden age of capitalism”, affluence in developed societies has been channeled towards physical goods. Consumer wealth, whether measured in time or money, would often be spent on better housing, tastier or higher-quality food, nicer cars and clothing, or an accumulation of objects or experiences to make our life more comfortable. Relating to Maslow’s hierarchy of needs, access to those consumer goods often matched the lower two levels of the pyramid i.e., physiological and safety needs.

As consumers had those needs covered, brands would strive to build aspirational value into their goods by associating them to the concept of prestige, success, and recognition, and hence targeting the higher levels of the pyramid: belongingness, esteem, and self-actualization. This aspirational value was communicated through media, and amplified through social interactions i.e., through the opinion that others would develop about us when we bought such goods. Media had already reached mass scale: TV, radio and papers could reach millions of potential consumers at once. However, social interactions were still limited in the sense that one’s social circle remained limited to family, friends, neighbors, or colleagues – a few dozen people, or at best low few hundreds.

The technological explosion of the early 21^st century, with widespread internet access first, the appearance of social media second, and the wide adoption of smartphones third, drastically altered these dynamics in at least two fundamental ways. First, unlike in prior consumption patterns, consumers of social media did not crave for physical goods, but mostly for emotional goods that matched the higher three levels of Maslow’s hierarchy of needs. They no longer sought to improve their physical living conditions, but craved for relationships and friends (belongingness), “likes” and recognition from others (esteem), and ultimately followers and success (self-actualization).

Second, technology-enabled social media eliminated barriers to the creation and distribution of content. It was no longer brands (through media) or a limited number of our social acquaintances who influenced our consumption patterns. Suddenly, the definition of what is desirable could be in the hands of pretty much anyone, and endorsed by little-known algorithms which often reward user engagement i.e., dopamine-based reactions. As social media users, we are more susceptible than ever to others mandating what is desirable to us.

3. Enter: Charlie Munger

Charlie Munger was an American businessman, investor, and philanthropist, and Warren Buffet’s longtime partner at Berkshire Hathaway[2]2. Munger died in 2023, barely a few weeks before turning 100. He is credited for his wisdom, and his ability to grasp the complexity of business and human behaviors and explain them in simple terms. One of my favorite Munger’s insights is that “the world is not driven by greed; it’s driven by envy”. Greed means everyone wants to have more, whereas envy means everyone wants to have more than their neighbor, no matter how much they already have. The prevalence of envy over greed has existed long before our modern society (ever heard of Cain and Abel?) and is deeply ingrained in nature, not only in humans but in other animals. If you can spare 2’43’’, this is a hilarious video showing a live example of animal envy.

My conviction is that the explosion of social media amplifies the impact of Charlie Munger’s words to a level not seen before. To the extent that social media multiply an individual’s reference points by enlarging our social networks well beyond our close circle, we now have many more people to compare ourselves against, and hence a much greater chance of being envious. On top, we perceive those comparison points as much more accessible: if my connections (and no longer a distant celebrity) have something or can do something, why can’t I? Or even worse: why shouldn’t I? The positive reinforcement mechanism that comes from creating social connections turns negative when something relevant (“I like it”) becomes desirable (“I want it”), we then develop a sense of entitlement (“I deserve it”), and ultimately get frustrated when we can’t achieve it all (“why don’t I have it?).

Envy, which in the old world was slower to spread because of the human-scale of physical networks, spreads faster empowered by virtual networks. To a certain extent, we lose pleasure from the process of achieving something, and get pleasure from shots of dopamine linked to instant achievement, as we are first flooded and then fooled by a ton of stimuli that seem easily accessible. Coming back to Anna Lembke’s thesis, the abundance of stimuli makes us less sensitive to them and therefore crave for even dopamine shots as compulsive social media consumers.

4. Impact on relationships

The Harvard Study of Adult Development is probably the longest study on adult life, having run for over 85 years. I strongly encourage you to watch this 13-minute TED talk about it, but if you don’t have the time (or if you also crave for low effort and instant achievement 😉), the TLDR is that good relationships keep us happier and healthier.

The good news is that both technology and social media are enablers for relationships. They help overcome physical distances and easily connect with each other. Facebook’s original mission was “to give people the power to share and make the world more open and connected”: a very noble goal, definitely worth pursuing.

The bad news is that amount and speed of connections are not synonyms of quality: it is not abundant and fast relationships that keep us happier and healthier, but good relationships that do. And for the reasons mentioned above, we’re often pushed to many and fast, over good.

At human scale, relationships and trust take time and effort: you rarely become friends with someone instantly. If you have read The Little Prince, you may recall the conversation between The Little Prince and the fox:

"I am looking for friends. What does that mean--'tame'?"

"It is an act too often neglected," said the fox. It means to establish ties."

"To establish ties?"

"Just that," said the fox. "To me, you are still nothing more than a little boy who is just like a hundred thousand other little boys. And I have no need of you. And you, on your part, have no need of me. To you, I am nothing more than a fox like a hundred thousand other foxes. But if you tame me, then we shall need each other. To me, you will be unique in all the world. To you, I shall be unique in all the world."

And later on, it continues:

"Please--tame me!" he [the fox] said.

"I want to, very much," the little prince replied. "But I have not much time. I have friends to discover, and a great many things to understand."

"One only understands the things that one tames," said the fox. "Men have no more time to understand anything. They buy things all ready made at the shops. But there is no shop anywhere where one can buy friendship, and so men have no friends any more. If you want a friend, tame me…"

"What must I do, to tame you?" asked the little prince.

"You must be very patient," replied the fox. "First you will sit down at a little distance from me--like that--in the grass. I shall look at you out of the corner of my eye, and you will say nothing. Words are the source of misunderstandings. But you will sit a little closer to me, every day…"

Strong relationships are patient, go through ups and downs, and require empathy and communication. And communication is not just words. Mehrabian’s rule establishes that in communications of feelings and attitudes, 7% of communication is through spoken words (content), 38% is through vocal elements such as tone, pitch, and inflection; and 55% is through non-verbal cues like facial expressions and body language.

Technology hugely facilitates relationships by shortening time and distances (calls instead of letters) and adding context (video instead of voice). But also, technology-based relationships can miss a lot of the communication cues that allow to build long-term trust and connection: discussions replaced by mails lose non-verbal communication, and mails replaced by chats further reduce depth. Speed, immediacy, and brevity generate impatience, which again is the opposite that we need to build good relationships.

The below chart shows the evolution of social media penetration (calculated as social media user identities3 as a % of world population4), superposed with the positive and negative experience index as defined by Gallup5. To be clear, the chart does not show neither correlation nor causation among them. It shows the explosion in social media penetration, a worsening in negative emotions, and a very slight (but almost flat) improvement in positive emotions. While there are certainly many factors that drive emotions, I remain intrigued by the extent to which dependency on social media contribute to the increase of such negative emotions.

5. The memecoin anecdote

In this context, some weeks back ago I came across a podcast on crypto and memecoins, where the guest interviewed a crypto guru named Murad Mahmudov. Although the podcast has in my opinion an awful clickbait title (“Get FILTHY RICH In The Year of Memecoins (2025)”), I still found it a worthy 1h10min learning time investment. While listening, I got somewhat struck by the following dialogue from min. 35:

Host: “I suffer from loneliness, or suffered pre-crypto a lot. Tell me how you think memecoins could solve loneliness”.

Guest: “I wouldn’t go as far as to say they’re going to solve loneliness completely, but loneliness is one of the 2025 forces that are driving people to crypto overall but specifically to memecoins and I think it might be a partial solution to these sorts of emotional voids that people have kind of in their hearts and souls. […] People find an escape in technology; […they…] are looking for places to belong both online and offline […] I think this phenomenon is happening in finance as well. Some of these dynamics are going to spill into crypto as well”.

The mechanism by which people struggle to build relationships, hence feel lonely, hence actively seek for belonging in a community, and how that community then gets “tokenized” through a memecoin whose value raises with community engagement, simply blew my mind. A lot of it may just be driven by opportunistic speculation, but still truth is that as I write this post the market cap of memecoins is approximately $48B6 – equivalent to the market cap of Germany’s Deutsche Bank7. A whole economy (partially) fueled by a need for relationships and belonging.

6. The age trap

Is it possible that all the above interdependent relationships between dopamine, relationships, and general happiness affect some more than other? My believe is they do.

Erik Erikson (1902-1994) was a psychologist who developed one of the most popular and influential theories of development, according to which humans go through eight psychosocial stages from infancy to old age8. At each stage, individuals face a core conflict between biological and sociocultural forces, which influences their personal and social development. If they are successful in resolving the conflict, individuals come out of that development stage with a healthier psychological outcome i.e., a “virtue”. Specifically, between the approximate ages of 6-11 children develop their sense of competence, acquiring new skills, building their confidence, and figuring out what they are good at. It is the conflict between industry understood as one’s ability to develop and master their skills to achieve something, and inferiority. In a successful outcome, individuals coming out of this this life stage having developed industriousness and confidence. In the following stage (ages 12-18) individuals explore their independence and develop a sense of self. It is shaped by social relationships, and reflects the conflict between identify and role confusion: successfully completing this stage leads to stronger commitment, self-confidence, sense of independence, and fidelity. Not doing so can result in “being unsure of who you are and where you fit, drifting from one job or relationship to another, or feeling disappointed and confused about your place in life”9.

While social media appeared in various flavors from the early days of the Internet, especially as chatrooms, their explosion started in the early 2000’s especially as users were allowed to share content through, among others, Friendster (2002), MySpace (2003), LinkedIn (2003), Facebook (2004), YouTube (2005), and Twitter (2006). First adopters were naturally younger users: according to the Pew Research Center10, in 2009 the percentage of online US teens that used social media was approx. 73%, similar to young adults aged 18-29 (72%) but much higher than adults aged 30+ (40%).

When social media becomes a major influence to our social interactions, it’s reasonable to expect that they also influence the outcome of those stages of development. For those which have made social media a core component of their childhood and teenage social interactions, their industriousness, confidence, identity, fidelity, and ability to commit (among other) will be influenced by typical social media dynamics like easiness of access, brevity, and impulsiveness, as opposed to in-person, real-life dynamics often characterized by longer trust-building cycles not exempt of conflict and ups-and-downs.

7. Wrapping up

Are we doomed? No.

Should we ban or punish social media? No.

What we likely need is not prohibition, but calibration. An honest look at how the tools we’ve built interact with the architecture of our minds and the timelines of our development. Not all innovation is neutral.

If memecoins, like memes themselves, are a cultural mirror —at times absurd, hyperbolic, and occasionally revealing—then perhaps they’re telling us something important. Not just about money, but about the emotional currencies we trade in: belonging, recognition, meaning. The question is not whether we should abandon the digital world, but how much of our humanity we are willing to outsource to it. We need to make that balance evident and actionable not just for those who already have accumulated life experience, but also who, because of age or other reasons, do not yet have that perspective.

Regardless of the outcome (spoiler: it worked!), I just loved the experience. To me, coding felt like solving a problem by breaking it down into elementary blocks that had to seamlessly fit together using efficient instructions. It combined problem-solving, logic, and tinkering – what else could an engineer ask for? But, for some reason, I never sustained that early passion after that summer: we didn’t have coding courses in subsequent years at university, Turbo Pascal was already pretty much obsolete (why the university taught an outdated coding system would make for a different post), and I didn’t have the time or patience to learn a more relevant language (C++) on my own.

Since then, the only time I’ve done anything close to coding was in my early years at Amazon (2014-2016), building occasional SQL queries to extract data and run basic business analysis. I still liked the logic involved in building queries and the underlying problem-solving it required, but I must admit it was less fun than I remembered from 1997. The context was radically different: there was a tighter time constraint which did not exist during my 1997 holiday, and a much larger knowledge gap to fill, both in terms of SQL language and the underlying data tables. This only increased my FOMO about whether I should (re)learn to code, at least as a hobby or maybe to keep up with an increasingly tech-driven world.

If any of the above resonates with you, I have some good news. Let’s fast forward to 2024.

Encouraged by a recent Amazon all-team event, conducted under the guiding principle of the company’s “Learn and Be Curious” leadership principle, and which included inspirational examples of how various Amazon teams had used (company-approved) GenAI tools to boost productivity, I felt it was time to overcome my personal FOMO about coding. I wanted to test how easy or hard it would be to build an executable piece of software (a game, in this case) using GenAI tools like ChatGPT or PartyRock.

The game is simple: as a player, your goal is to click inside a square that randomly moves every two seconds. To make it harder, as soon as you hover the mouse over the square, a new timer kicks in, and the square moves only a fraction of a second later, without waiting for the end of the full two seconds. If you're fast and manage to click on the square, you get a point. If you're not fast enough and miss it, you get no points and need to chase the square again to its new position. Total play time is 60 seconds.

Given my zero knowledge about coding (whatever I learned about Turbo Pascal 27 years ago is now both forgotten and irrelevant), building such an app from scratch would have taken me weeks. But with GenAI tools, it took me just three hours from start to finish, including all the inefficiencies along the way. Here’s how the process unfolded.

I started with PartyRock, using a very simple prompt for a basic version of the game: “I'd like to do a game. The goal of the game is to be able to move a pointer inside a square of 2cm x 2cm, and click inside it. However, to make it difficult, the square will change position in the screen 0.2 seconds after the pointer has reached inside it before clicking”. That prompt generated an interactive HTML5 canvas, a few plain-language code explanations, and some game instructions. Since I had no idea what an “interactive HTML5 canvas” was, I asked ChatGPT (“I have no idea of programming, but have generated an HTML5 canvas game. How can I try it?”), and got a simple answer explaining what it was and how to get the game running. The game wasn’t exactly what I’d asked for, as the square wouldn’t move at all until clicked on, but it was still impressive for just a few minutes of effort.

From there, getting to a finished version of the game took nine more iterations. I switched to ChatGPT 4o, as I found it easier to build on prior prompts. These iterations added features like a score counter, a success jingle for each point, a timer, and three levels of game complexity. They also helped with debugging, as the square would sometimes move before the two-second timer finished, even without a click. Debugging was probably the most frustrating part of the experience: I felt repeatedly misunderstood even if I described ChatGPT what was not right. It was just like when you try to get something fixed, and the person on the other side ignores what you want or simply does not get it:

• “[…] The square movement is still not right. When the player successfully clicks in the square, the square will move (which is OK), but then will move again before the new, re-set 2-second timer is finished. That's not right. Please correct.”

• “The issue remains. Please re-check!!!”

• “Always the same problem. Please fix.”

Still, the larger chunk of the three hours was not lost in asking ChatGPT to debug, but rather in myself trying to read and understand the code, hoping that I would be able to debug myself. In hindsight, that was useless, inefficient, and wrong.

Needless to say, when the game finally worked, I was both amused and amazed. Amused because it had been a fun experience that reconnected me with the 1997 Tetris exercise. Amazed because of how quickly it all came together. It also sparked many questions about the future of coding. Chief among them: Is coding dead as a discipline?

I don’t know, but I don’t think so. Building my game was a very simple, isolated exercise. I can’t imagine an AI tool generating complex, high-value software based on prompts from a tech-illiterate person and having that result in a widely successful product. Not in the short or mid-term, at least. The large number of dependencies between disparate systems, increasingly complex security requirements, and the human need to “feel in control” make it unlikely that we’ll leave software building entirely in the hands of machines anytime soon. Even if counterexamples exist, I believe they’ll be the exception rather than the rule—they won’t be widespread enough to kill coding as a discipline. Instead, as I mentioned in this previous post, mediocre coders will have a harder time. They’ll face competition from less experienced but curious and resourceful individuals who can build their own applications with GenAI tools. Meanwhile, strong, experienced, high-judgment coders will boost their productivity further by using these same tools.

I’m clearly not the first person to try “coding” with zero knowledge using GenAI tools. The web is filled with examples (here’s one by tech entrepreneur Martin Varsavsky). If you haven’t tried it yet, I encourage you to do so, even with a simple application. In the meantime, if you’d like to test the game I built, below is the code. You can copy-paste it into a text file (txt), change the extension to .html, and double-click to play.

Have fun (non-)coding!

<!DOCTYPE html>

<html lang="en">

<head>

    <meta charset="UTF-8">

    <meta name="viewport" content="width=device-width, initial-scale=1.0">

    <title>Click the Moving Square Game</title>

    <style>

        body {

            margin: 0;

            display: flex;

            justify-content: center;

            align-items: center;

            height: 100vh;

            background-color: #f0f0f0;

        }

        canvas {

            border: 1px solid black;

        }

        .score-timer {

            position: absolute;

            top: 10px;

            left: 10px;

            font-family: Arial, sans-serif;

            font-size: 20px;

        }

    </style>

</head>

<body>

    <canvas id="gameCanvas" width="800" height="600"></canvas>

    <div class="score-timer">

        <span id="score">Score: 0</span> |

        <span id="timer">Time Left: 60</span>

    </div>

    <!-- Success sound -->

    <audio id="successSound" src="https://www.soundjay.com/buttons/button-11.mp3" preload="auto"></audio>

    <script>

        const canvas = document.getElementById('gameCanvas');

        const ctx = canvas.getContext('2d');

        const successSound = document.getElementById('successSound');

        const squareSizeCm = 2;  // Size in centimeters

        const pixelsPerCm = 37.7952755906;  // 1 cm in pixels

        const squareSizePx = squareSizeCm * pixelsPerCm;  // Convert size to pixels

        let squareX = Math.random() * (canvas.width - squareSizePx);

        let squareY = Math.random() * (canvas.height - squareSizePx);

        let score = 0;

        let timeLeft = 60;

        let gameTimer, moveTimer, mouseOverTimer;

        let mouseOverDelay = 500;  // Default for Beginner: 0.5 seconds

        let gameRunning = true;

        // Ask player for complexity level before starting the game

        function chooseLevel() {

            const level = prompt("Choose your level: 'Beginner', 'Intermediate', or 'Expert':").toLowerCase();

            switch (level) {

                case 'beginner':

                    mouseOverDelay = 500;  // 0.5 seconds

                    break;

                case 'intermediate':

                    mouseOverDelay = 350;  // 0.35 seconds

                    break;

                case 'expert':

                    mouseOverDelay = 150;  // 0.15 seconds

                    break;

                default:

                    alert('Invalid choice! Defaulting to Beginner.');

                    mouseOverDelay = 500;  // Default to 0.5 seconds for Beginner

            }

            startGame();  // Start the game after selecting the level

        }

        // Draw the square

        function drawSquare() {

            ctx.clearRect(0, 0, canvas.width, canvas.height);

            ctx.fillStyle = 'red';

            ctx.fillRect(squareX, squareY, squareSizePx, squareSizePx);

        }

        // Move the square to a new random location

        function moveSquare() {

            clearAllTimers(); // Ensure all timers are cleared before moving

            squareX = Math.random() * (canvas.width - squareSizePx);

            squareY = Math.random() * (canvas.height - squareSizePx);

            drawSquare();

            resetMoveTimer();  // Reset the 2-second movement timer after each move

        }

        // Clear all timers (moveTimer and mouseOverTimer)

        function clearAllTimers() {

            clearInterval(moveTimer);

            clearTimeout(mouseOverTimer);

            mouseOverTimer = null;  // Ensure mouseOverTimer is also reset

        }

        // Update score display

        function updateScore() {

            document.getElementById('score').textContent = `Score: ${score}`;

        }

        // Update timer display

        function updateTimer() {

            document.getElementById('timer').textContent = `Time Left: ${timeLeft}`;

        }

        // Start the game

        function startGame() {

            drawSquare();

            resetMoveTimer();  // Start the 2-second movement timer

            gameTimer = setInterval(() => {

                timeLeft--;

                updateTimer();

                if (timeLeft === 0) {

                    endGame();

                }

            }, 1000);  // Decrease time every second

        }

        // End the game

        function endGame() {

            clearInterval(gameTimer);

            clearAllTimers();  // Clear all timers

            gameRunning = false;  // Prevent further game actions

            alert(`Time's up! Your score is: ${score}`);

        }

        // Reset the 2-second movement timer

        function resetMoveTimer() {

            moveTimer = setInterval(() => {

                moveSquare();

            }, 2000);  // Set a new 2-second timer

        }

        // Check if mouse is inside the square

        function isMouseInsideSquare(mouseX, mouseY) {

            return mouseX >= squareX && mouseX <= squareX + squareSizePx &&

                   mouseY >= squareY && mouseY <= squareY + squareSizePx;

        }

        // Handle mouse movement over the canvas

        canvas.addEventListener('mousemove', (event) => {

            if (!gameRunning) return;  // Do nothing if the game has ended

            const rect = canvas.getBoundingClientRect();

            const mouseX = event.clientX - rect.left;

            const mouseY = event.clientY - rect.top;

            if (isMouseInsideSquare(mouseX, mouseY)) {

                if (!mouseOverTimer) {  // If there isn't already a mouseOver timer running

                    mouseOverTimer = setTimeout(() => {

                        moveSquare();

                    }, mouseOverDelay);  // Use the selected delay (0.15/0.35/0.5 seconds)

                }

            } else {

                clearTimeout(mouseOverTimer);  // Cancel any pending mouse-over movement if the mouse leaves the square

                mouseOverTimer = null;

            }

        });

        // Handle mouse clicks

        canvas.addEventListener('click', (event) => {

            if (!gameRunning) return;  // Do nothing if the game has ended

            const rect = canvas.getBoundingClientRect();

            const mouseX = event.clientX - rect.left;

            const mouseY = event.clientY - rect.top;

            if (isMouseInsideSquare(mouseX, mouseY)) {

                score++;

                updateScore();  // Update the score on successful click

                successSound.play();  // Play success sound

                moveSquare();  // Move the square immediately after the click

            }

        });

        // Prompt player to choose their level before starting the game

        chooseLevel(); 

    </script>

</body>

</html>

Here they are without further introduction and in chronological order of each production. I look forward to reading comments about other examples that have left a mark on you and which may make for a good movie or TV plan!

1. A Few Good Men: facing the truth even when it’s difficult

In the drama "A Few Good Men" (1992), a young Tom Cruise plays an inexperienced U.S. Navy lawyer looking to shed light on the death of a Marine, victim of what looks like an extrajudicial punishment carried out by someone at his own base but ultimately commanded by Base Commander Colonel Nathan Jessep (character played by Jack Nicholson).

In the final courtroom scene, Tom Cruise’s character seeks to corner Jack Nicholson's character into responding about his involvement in the death, which he resists until delivering the iconic line:

-         You want answers?

-         I want the truth!

-         You can’t handle the truth!!!

I bring up this scene because leaders will often see themselves on both sides of the dialogue: facing hard truths, and deciding whether and how to communicate on them. They need to be ready to hear and cope with a reality they don’t like or expect, but they’ll also have to choose the level of information and transparency they want to share with their organizations: downward to their teams, sideways to their peers, or upward to their managers. How much information can you share? Can your organization ‘handle the truth’? When are you too transparent?

When on the receiving side, my firm conviction is that you’re better off facing reality no matter how hard it is: that will make you better equipped to resolve any downstream issues that you’d otherwise ignore at your own peril. In the cartoon below1, and even more so if you’re the leader of an organization, you want to line up next to the emptier table.

When on the giving side, effective leaders understand the importance of transparency and honesty towards their teams: while it may be uncomfortable to deliver bad news or admit mistakes, concealing the truth can be even more damaging. Trust is the foundation of any successful organization, and trust can only be built and maintained when leaders are forthright with their teams and stakeholders. Ever heard of Churchill’s Blood, Toil, Tears and Sweat speech, or of the Streisand effect? If an organization “can’t handle the truth”, that’s a sign that there’s a flaw somewhere in it, in its values, or the engagement and maturity of its employees, no matter at what level. BUT we must also acknowledge that the real world is not as we’d like it to be, and an excess of transparency, e.g., when sharing bad news, can well create doubt, disengagement, or anger within the organization. To avoid that, leaders need to realize that delivering bad news isn't just about sharing information; it's about providing context, solutions, and a path forward. Leaders must be prepared to support their teams through difficult times, offering guidance and reassurance. In doing so, leaders not only demonstrate their commitment to the truth but also foster a culture of resilience and adaptability.

2. Spider-Man: "With great power comes great responsibility"

This quote, which some of us may remember from the 2002 film "Spider-Man", was already introduced by Spider-Man co-creator Stan Lee in a 1962 edition of the comic book Amazing Fantasy. Clearly not a novel idea—it is, after all, the central idea behind the story of the Sword of Damocles, which dates back to the 1st century BC—but one that was crisply and neatly captured in the superhero story.

As we grow in our careers, we often aspire to higher positions of power, but those are often ill-perceived as positions where we will enjoy higher autonomy to make decisions, less operational friction to get things done, and often a team to do the work. Or, in plain terms, where we will be able to decide, command, and harvest results without having to get our hands (too) dirty with execution. Yet, reality is way more complex and not quite as favorable.

In business, power comes in many forms—the power to make decisions, allocate resources, and influence others. But those powers come with a number of obligations: to quote a few, the need to rally others around such decisions, earn their engagement, resolve conflict, and accept the consequences of our decisions, whether good or bad. Power without responsibility is fundamentally flawed: commanding top-down may work for some brief period of time, but it will be a short-lived leadership style, as are leaders who frequently blame those under their responsibility. As leaders ascend to positions of greater authority, they must never forget their responsibility to their teams, organizations, and stakeholders. By embracing that ethos, they can foster trust, loyalty, and a sense of purpose among their employees, ultimately multiplying the positive impact of their leadership.

3. Zero Dark Thirty: embracing ownership

“Zero Dark Thirty,” a 2012 thriller directed and co-produced by American filmmaker Kathryn Bigelow, dramatizes the decade-long hunt for Osama bin Laden, leader of Al-Qaeda. One of my favorite scenes in the movie depicts a CIA meeting in which the lead is scolding a team of ~15 analysts, among whom lead character Maya (played by US actress Jessica Chastain). The monologue goes something like this:

“I wanna make something absolutely clear. If you thought there was some secret cell somewhere, working Al-Qaeda, then I want you to know that you’re wrong. This is it. There’s no working group coming to the rescue. There’s nobody else hidden away on some other floor. There is just us. And we are failing.”

Leaving aside the tone and (needless to say) the context, what I like about this scene is the sense of ownership it conveys.

Especially in large organizations, it is tempting to think that there’s built-in redundancy. That if we do not get to do something, do it late, or deliver a substandard output because of failure or procrastination, someone else will pick that up behind us: a colleague, a team member, or even our manager. Not only is that false, but such dilution of ownership is hugely damaging to the organization.

It's essential to recognize that leaders can't always rely on someone else to come to the rescue. In many cases, they are the ones with the unique skills, knowledge, and perspective needed to navigate through challenges successfully. But even if they’re not, taking ownership of a difficult or even boring task is a sign of leadership—irrespective of whether you hold a manager or an individual contributor role in your organization. This willingness to lead by example inspires confidence in the broader team and fosters a culture of accountability and determination.

4. The Diplomat, on delegation

My last example is not from a movie, but from a great TV show called "The Diplomat" (2023). In season 1, the main character (Kate Wyler, played by actress Keri Russell) is newly appointed as US ambassador to the UK against her will and soon needs to navigate first-hand a series of political and geopolitical escalations. In Episode 5 of Season 1, the following dialogue takes place between Ambassador Kate Wyler and her deputy chief of mission or DCM (played by US actor Ato Essandoh). In a moment when she’s running around going after some information she wants to get by herself, he challenges her:

[DCM]. Hi! You know I work for you, right? As does Ronnie?

[Ambassador]. Yeah

[DCM] You were a DCM, but you’re not anymore. So, if you need to figure stuff out, I am a resource. As is Ronnie, as are the 800 other folks that work for you.

[Ambassador] By the time I explain it…

[DCM] Madam, I think you should try. I think that we are doing the part of the job that you like, you’re great at it, but it is a big job. So is mine. You can’t handle both.

The above dialogue summarizes one of the most common errors in leadership: the inability to delegate. In its mildest consequence, it results in overwork for the leader. In a worst case, the inability to delegate creates a disjointed organization where leaders become bottlenecks that cripple growth, and those in lower ranks don’t get the chances to develop and grow.

The risk of this happening is greater when the leader has risen from the ranks, knows how to do the job of his/her employees, or finds deep satisfaction in showing expertise on the topic. Or when he/she shows excessive ownership, i.e., everything becomes not just his/her responsibility but his/her job. A lot has been written about effective delegation, and I won’t try to summarize it here. But the extract above is a great reminder that as we grow in our careers, we also need to learn how to ‘let things go,’ trusting our teams to take care of them, even when it feels as a short-term cost, for it is a worthy long-term investment.

These scenes serve as helpful reminders of some important leadership traits: the commitment to truth and transparency, the responsibility that comes with power, the need to create a sense of ownership in the organization, and the need to rely on teams to reach higher grounds. Again, I look forward to reading comments with additional suggestions of leadership examples from other movies or shows!

Leaving aside the technicalities of Huang’s speech (e.g., what a NIM is), the above made me reflect on how I had always thought that Artificial Intelligence would materialize in sophisticated but independent software applications. In reality AI is likely to be subject to similar dynamics as human intelligence. This means there might not be one AI but different types of Artificial Intelligence, and it will not materialize in isolated applications but in larger systems where individual applications perform specific functions, effectively resulting in an AI division of labor. It is a revised notion which drastically expanded my current understanding of artificial intelligence and drew a parallel with the multifaceted and complex nature of human intelligence and labor division.

Looking at how human intelligence manifests, one of the most widely recognized frameworks is the one that Howard Gardner introduced in his 1983 book Frames of Mind: The Theory of Multiple Intelligences. Gardner established that humans have various kinds of intelligences, each of them associated with different abilities:

1. Linguistic Intelligence: The ability to use words effectively, whether orally or in writing.

2. Logical-Mathematical Intelligence: The capacity to analyze problems logically, carry out mathematical operations, and investigate issues scientifically.

3. Spatial Intelligence: The ability to think in three dimensions.

4. Musical Intelligence: The capability to recognize and compose musical pitches, tones, and rhythms.

5. Bodily-Kinesthetic Intelligence: The capacity to use one's whole body or parts of the body to solve problems or create products.

6. Interpersonal Intelligence: The ability to understand and interact effectively with others.

7. Intrapersonal Intelligence: The capacity to understand oneself, to appreciate one's feelings, fears, and motivations.

8. Naturalist Intelligence: The ability to recognize, categorize, and draw upon certain features of the environment.

9. Existential Intelligence (a ninth proposal by Gardner), involving the ability to tackle deep questions about human existence, such as the meaning of life, why we die, and how we got here.

Different types of intelligence mean people are skilled in different fields, which in economic and professional terms make them also more productive for specific jobs (this article offers a simplified approximation based on Gardner’s framework above). When these skills come together in a complex human system (society), they naturally enable a division of labor which in turn brings prosperity. Quoting the book Abundance: The Future Is Better Than You Think, it “allows specialists to exchange specialties, thus creating what economist Friedrich Hayek called catallaxy: the ever-expanding possibility generated by the division of labor”1.

Unlike human intelligence, artificial intelligence today seems to be most commonly classified2 based on its level of sophistication, and not yet or not as much on the type of intelligence it demonstrates. This way we distinguish between:

• Narrow or Weak AI (ANI), designed to perform a narrow task (e.g., facial recognition, internet searches, or driving a car). This is the type of AI that we interact with in our daily lives via for example Alexa or Siri, or through recommendation systems like those used by Amazon or Netflix.

• General or Strong AI (AGI), which would have the ability to understand, learn, and apply its intelligence to solve any problem, similar to how a human would. This form of AI does not yet exist and is considered a goal for future AI research.

• Superintelligent AI (ASI), a hypothetical form of AI that surpasses the brightest and most gifted human minds in practically every field.

What the earlier quote from Jensen Huang suggests is that AI applications are not to be thought in isolation, nor measured just by their level of sophistication (ANI vs. AGI vs. ASI). Just as humans interact in a complex societal system and play a somewhat specialized role based on their skills in the larger division of labor of the economy, AI applications should also be looked at systemically.

Hayek's concept of catallaxy provides a compelling framework for envisioning the future of AI, too. In a catallactic AI ecosystem, various AI entities, each with its unique capabilities, could autonomously collaborate, innovate, and contribute to a harmonious order not centrally directed but collectively beneficial. Different AI applications will stand out for their different strengths or skills, will add value at distinct spots along the value chain of a company or the broader economy, and will interact with each other in order to deliver an output that is greater than the sum of their individual abilities. Just as skills and division of labor exist in human-based systems, one day we shall see similar distinct skills and division of labor in fully AI-based systems.

This model hints at a future where AI systems not only perform tasks but also engage in decision-making and resource allocation, perhaps even developing a new economy of AI-driven value exchange. But… what will those AI systems look like, and how will they govern themselves? For example, will hierarchies emerge in AI-based systems in the same way as they have emerged in human ones? Will there be “manager AI applications” which will govern the inputs and coordinate the outputs of individual, lower-level AI applications? Under what rules? And to the extent that such AI systems will require individual AI applications to interact (or even “trade”) with each other based on their skills, will there be a different concept of value or price governing those exchanges to the one we have known until now?

In the words of Huang, we may not be writing software from scratch in the future, but rather orchestrating various AI talents towards a shared purpose. The potential for AI to mirror the division of labor and specialization seen in human society opens new avenues for prosperity and innovation. However, the autonomy of AI systems also raises critical questions about their alignment with human values, the balance between automation and human labor, and the mechanisms to address inequalities stemming from differential access to AI technologies.

If seeing the development of (individual) AI applications is already a fascinating journey, imagining how AI will evolve from individual applications to large and complex AI systems is nothing short of mindblowing.

Both my brother and I have been lucky enough to accompany you for most of these 50 years. My goal today is hence to remember the mark that you have left on us during this time, and to thank you for it. Starting with the values that you have transmitted to us.

You have instilled in us that you are happier when, instead of having what you want, you simply want what you have, and you have what you need. That we must always do the right thing, at all times, even if it is not what we like the most or it is not the easiest path. Because by doing the right thing, even the most complex and ambiguous decisions become simple.

You have taught us the value of personal effort. To give our best version at all times, without fuss, without exaggerations, and without that effort exhausting our energy. To appreciate a job well done, and to not settle for delivering less than what you are capable of, even if the reward does not come in the short term.

You have also taught us that not everything is numbers, but that above numbers are people. I admit that, as an inveterate engineer, this was a little more difficult for me to understand and apply. But since patience is another of your virtues, and another of the values that you have transmitted to us, you knew how to insist and wait for the message to sink in.

You inherited many of these values from our grandparents, whom we remember so many times. I suppose that one of the greatest aspirations and satisfactions of a parent is to be proud of her children, and I know that they are proud of you seeing what you have created.

A couple of days ago I heard someone say that loving is caring, understanding, and inspiring. For us, you have achieved all three perfectly. You have cared for us at all times, and sacrificed your own well-being so that we could achieve ours. You have understood us in our particularities and in our life choices: our studies, our trips, our families. And you have inspired us to, as I said before, achieve the best version of ourselves.

We are a family sparse in demonstrating feelings, but also a family that appreciates the small details. Furthermore, love often manifests itself thus, in small details. That's why I also want to pay tribute to those small details that, going almost unnoticed, show the love you have for each other. Like the fact that you still walk hand in hand (not arm in arm, hand in hand) after so many years. That there’s no meal that dad doesn't thank and praise with a “it's delicious, love”. Or that, even when you have to tell each other that they are making a mistake, you do it calmly; many times, simply resting your hand on his arm or his knee. And just with that, the other person knows that he/she has to hold himself or herself back.

With all this, I can only thank you on behalf of my brother and me. Thanks for the playing time: the towers made of yogurt cups, the marbles, the bottle caps. Thank you for the tupperwares, which continue to flow even well into our 40s. Thank you for bringing us all together in family celebrations, and for joining us in our celebrations even when they were far away and you had to catch a plane. Thank you for the education you have given us – I can't imagine a better one. Thank you for supporting us in our decisions, even when sometimes you certainly did not understand them, or sometimes even did not share them. Thank you, after all, for loving us so much.

We love you.

My original intent when writing this article was to foresee, maybe guess, which jobs would be last to be replaced by AI, and hence ‘safest’ from automation. But as I wrote down and checked my own hypothesis, I could pretty much find counter-examples for each of them. So – and this is a spoiler – this article is eventually to state that I simply have no idea. If you expect a clear and definite answer, do yourself a favor, spare yourself ten minutes of reading, and move on to something else. Similarly, if you are looking for yet another apocalyptic article in the style of ‘AI will replace us all in our jobs’, please stop reading as well. However, if you also ask yourself which jobs are safe(r) from automation, and are interested in reading some hypothesis and counterexamples, then keep going. I would love if you even leave comments about your own thoughts or critique those I’ve shared.

And now, let’s get to it.

1. The power of data (or absence thereof)

The Industrial Revolution and the automation that emerged from it mostly resulted in partial substitution of so-called blue-collar jobs: agriculture, manufacturing, construction, mining, or maintenance-related jobs3. From the beginning of time, and even more so from the creation of the steam engine and until (still) today, efforts have gone into creating machines that outperform human physical capabilities like speed, strength, or ability to perform repetitive actions with sufficient precision. Machines are faster to plow, dig, saw, or shape metals, but also perform precision activities like precision-cutting, sewing, welding, or engraving.

In the 20^th century, computers started performing not physical, but intellectual tasks such as complex calculations, organization of tasks, retrieval of information, or even entertainment. For a long time, though, those activities were subject to pre-programmed rules. Even if complex or intricate, computers would follow instructions defined as code, and execute accordingly. Given a number of inputs (data) and rules (code), the computer would generate an output, a result, an action.

AI changes that paradigm, in that it no longer outperforms human’s physical or intellectual capabilities that can be codified or programmed, but also those that can’t, such as abstract knowledge, inferences, and creative processes. In doing so, AI no longer relies on inputs and rules to produce results, but instead on inputs and results to create rules that can be applied for other scenarios, even different ones. If we were to describe it in human-like terms, AI is no longer an obedient person who performs actions based on orders and rules, but instead a reasoning person that can infer rules from the observation of external events and results – pretty much like we all learn in life. If conventional programming was based on the equation “Data + Rules = Results”, AI is based on the equation “Data + Results = Rules”.

But… what if there was no data to feed AI? Based on the above explanation, AI would be unable to thrive. If we extrapolated that same inference to the job market, we could think that jobs that don’t rely on data or for which there is no data readily available should be safer from AI-driven substitution. That is, for example, research and explorative jobs where by definition there is no prior data to rely on, creative jobs, or what I’ll call six-sigma jobs whose nature is to deal with extremely rare events that have little or no precedent (like a business leader who needs to cope with a severe, once-in-a-lifetime event, or an entrepreneur trying to prove the viability of a totally new idea).

However, truth is that all those jobs mentioned above do rely on some form of data – even if just hints of it. Research projects are often born from relationships between observations of one or several disciplines. Creative jobs also unavoidably take some inspiration from the past, even if only to deconstruct or adapt it. And leaders pull from past analogue experiences to try and navigate new circumstances. If you prompt ChatGPT, or Dall-E 2, or any other generative AI tool, to create a story or a picture for you, it will likely surprise you with an output you did not quite expect. If you ask it for how to cope with a once-in-a-lifetime event, it will probably suggest a course of action that is at least as good as the recommendation of an average person. Maybe not that of highly smart or resourceful experts, but good enough to outperform many of us. Which leads us to another hypothesis of AI-substitutable roles.

2. The dangers of mediocrity

McKinsey & Co. recently published a report named The economic potential of generative AI. The following paragraph particularly caught my eye:

“Research found that at one company with 5,000 customer service agents, the application of generative AI increased issue resolution by 14 percent an hour and reduced the time spent handling an issue by 9 percent. […]. Crucially, productivity and quality of service improved most among less-experienced agents, while the AI assistant did not increase—and sometimes decreased—the productivity and quality metrics of more highly skilled agents. This is because AI assistance helped less-experienced agents communicate using techniques similar to those of their higher-skilled counterparts.”

Imagine you’re the employer in that customer service center. Thanks to the use of AI, your less skilled and likely lower-cost agents can become as productive as your most skilled and also higher-cost ones. What would you do? The simplest answer might be to get rid of all but your less skilled employees, as they have now become highly productive. A more nuanced and likely righter answer would be to also keep your most skilled employees, to be able to deal with edge cases that AI might not be able to cope with.

But… what about those employees in the middle? They’re probably not distinctive enough to make a difference without AI, and are also more costly than their junior peers who suddenly have become equally as productive thanks to AI. That sounds like a tough spot to be in, right? Fortunately for them, there are many reasons why an employer would want to keep those middle-range employees, too. For example, building a pipeline of talent who can eventually become highly skilled (as current top employees may well rotate or leave at some point), repurposing them to another role where their skills make a greater difference, or simply maintaining a minimum workforce size to deliver on business commitments. But despite all these reasons, if you’re in that middle range, you have just become more replaceable than you were before.

It is likely that AI will boost productivity of less skilled workforce, and make it hard to succeed for ‘mid-skilled’ workers who can’t keep their edge in the new context, while highly skilled ones may spare themselves. You don’t want to be stuck in the middle.

3. De-scale

When I think about job automation, whether physical or intellectual, I tend to think about machines that are capable of outperforming humans because they can do the same tasks but faster, more precisely, or with greater strength. In other words, at scale. Earlier in this article I alluded to such examples: basic ones like plowing, highly precise manual tasks like sewing, or complex intellectual ones like calculations or financial modelling. For fun, I recommend you to watch this 14-second mesmerizing video of a machine triaging tomatoes, which shows how human capabilities can be dwarfed by the power of machines.

I sometimes convince myself that humans have a higher chance of outperforming machines and AI in high-variability jobs that require some degree of physical dexterity. As a very basic example, think of a waiter in a busy terrace: difficult to imagine a robot competing with the complexity and variability of tasks: take orders, serve and clean-up table after table, dodging an ever-changing set-up of tables, chairs, and patrons, and adapting to the needs of each one of them: whether a quick snack or a slower and more personalized service. Think also about the cook in that restaurant’s busy kitchen, having to deal with a vast range of menu combinations4, often customized to a customer’s needs, and at almost random timings which nevertheless need to be matched by guest table. Or to refer to a different industry, think of construction, where machines have for long helped increase productivity, but are not close to substituting basic manual jobs given the infinite variability of the outputs (whether houses, civil constructions, or any other type of building), and the physical dexterity needed to produce them. But despite my conviction, there are also counterexamples that show that AI can replace jobs exposed to high variability and which require physical dexterity: autonomous driving is such an exceptional use case, even if still in its infancy.

What makes some jobs replaceable, and some not, then? In my opinion, the differential value of human jobs is in their uniqueness, not on their scale. The higher the variability, and the lower the scale, the higher the personalization that a human can add vs. a machine.

4. Feelings

A few days ago, I came across a short video of Spanish comic Ignatius Farray, in which he claimed that “the last frontier beyond which we could really claim the concept of ‘Artificial Intelligence’ is when AI will feel shame, [as] shame is the purest human feeling”. I’d argue we can probably expand that to other human feelings: pity, envy, or greed, for example.

Machines have shown very capable of triggering or boosting such primary emotions in humans, but to my knowledge they have not demonstrated them, nor even credibly faked them. As my brother says, this is because we have a misconception of what AI is, and we call ‘intelligence’ what in reality are advanced statistical models – and a statistical model, no matter how advanced, will struggle to credibly express (not imitate!) feelings. Thanks to those feelings, I am convinced that humans can outperform machines in building relationships. I therefore believe that jobs that depend highly on creating, understanding, or managing relationships and emotions will be somewhat safer from full replacement by machines.

Think about management jobs, for instance. I’m convinced that machines can do a great job in organizing the work of a team, likely even better than a human can do. Machine algorithms can probably set priorities based on a number of well-quantified economical (or financial, or operational) goals and metrics, assign resources based on team members’ skills, prepare narratives to communicate, and of course track performance. But a core aspect of any management role is the management of feelings, especially those of team members, and not just the management of tasks. Being able to understand when a team member is going through a tough time and needs some flexibility or relief, or when a person might be ready to take a larger role (and whether he/she wants it or not!), or when a person needs some extra support to grow, are all emotional traits of management that machines will struggle to replace.

What about sales and commercial roles, which often rely on a lot one’s ability to understand and empathize with others’ feelings? I’m rather torn on this one. As Reid Hoffman explains in his book Impromptu: Amplifying Our Humanity Through AI, “AI will make human BDRs [i.e., Business Development Representatives] more effective by providing them with personalized information on the prospects they call – but this increased productivity will likely decrease employment. […] I believe that the future will see the sales profession shrink as a whole”. In many ways, machines have already gained their place in sales: for decades now, e-commerce sites have allowed brands to reach and sell items to large audiences which would be difficult to reach through a traditional salesforce. More broadly, machines can outperform humans in many steps of the sales process, e.g., scouting, targeting audiences, pricing, or building tailored proposals. I could even imagine for two parties in a negotiation to rely on machines (and not humans!) to close a deal: machines might be able to identify a Zone of Possible Agreement (ZOPA) where humans might fail because of a load of negative emotions like mistrust, envy, greed, or fear. But for that exact same reason I believe humans still play, and will always play in the future, a crucial role in sales. In the same way that negative emotions can make a deal fail, positive emotions can save or generate deals that are perfectly logical but would otherwise be dead. The best sales people excel at conveying trust, collaboration, generosity, happiness, or hope, all of which are deal facilitators. Who hasn’t bought something because the sales person called you by your name, had a kind word, or expressed empathy towards yourself as a client or towards someone who was with you?

Because of this, I am inclined to believe that machines (and among them, AI) will not replace jobs that rely on feelings. I mentioned earlier that the differential value of human jobs will be (or actually is) in their uniqueness. What’s more unique, and more human, than feelings?

5. Intent, curiosity, and ambition

Could AI ever desire to circumnavigate the world? Or to go to the moon, or to Mars? Or to explore the origin of life, or understand what happens after death (if anything)?

Intent, curiosity, and ambition seem human traits that a machine – even an intelligent one – could never replicate. The desire to explore something new, to try something for the first time, to experiment. For sure, machines (intelligent or not) can help in such physical and intellectual quests; they can allow humans to be faster in their research, safer in their efforts, or even to see things that might otherwise go unnoticed. But that initial spark, the direction, is human.

Certainly, AI (and even more so Generative AI) can take actions that look like intentional and self-directed in our day to day. If you hold a conversation with ChatGPT, or if you rely on an AI-generated recommendation for your next book, movie, or Instagram feed, you could argue that a machine is directing you rather than taking direction from you. But is that really the case? I’d argue not. While ML can be used for multiple problems, it is not to be used (among other) for problems where AI recommendations “require full interpretability”. Which somehow suggests that, in the earlier simple examples, AI-powered recommendations might be perfectly accurate, but somewhat unexplainable. Now, think about your school days, and imagine for a second that you had always provided the right answers to your teacher’s questions, but you were equally unable to explain why you got them right. That may not have gotten you too far, right? Areas where the why, the intent, the ambition, or the curiosity are important are, and I believe will be at least for some time, fertile territory for humans to outperform AI.

Or will they? Again as a counterexample, let’s look at what happened in 2016, in a series of Go games played between Lee Sedol, one of the world's best Go players, and Google’s AlphaGo technology. This is one of the first cases where AI demonstrated intentionality as a human would do, or at least so it seemed. As described in this article, with the 37^th move in the 2^nd game, AlphaGo took its opponent and spectators by surprise by making a move that many considered simply a mistake. I don’t play Go, so you’ll forgive me I can’t explain it to you in simple terms. But the important thing is that this one-in-ten-thousand move eventually proved key for AlphaGo to win the game, and according to the lead researcher on the AlphaGo project, David Silver, the machine had somehow viewed the move, and chose to execute it. Does that mean the machine demonstrated an actual intent with that move? Could be…

6. What will not change?

If you’ve read anything about AI, by now you have probably come across many articles anticipating or guessing the deep transformations that AI will cause in our lives – personal or professional. I have also alluded to a few changes in the paragraphs above. But quoting Jeff Bezos, it is even more important to ask ourselves what’s not going to change as a result of the growth of AI.

We, as humans, will still be moved by feelings, noble or not. We’ll remain social animals even if that may not be obvious when we spend hours in near-isolation, consuming information on our smartphones, tablets, or laptops. We’ll eat, get dirty, or get sick, and will need a way of solving for those (and other) physical needs. We’ll need education, skilling, and re-skilling all along our lives, to enjoy a minimum – or a maximum – of autonomy and independence. We’ll need to be entertained, protected, emotionally moved, and loved. We’ll need to be taken care of when we’re young, and when we’re old. We’ll be, after all, interdependent.

Learn to cater for those in a differential, human manner, and you’ll likely outperform AI.

A few months later, in February 2023, NBA commissioner Adam Silver presented a new NBA App which allows fans to insert their avatar in a live game to replace any field player, just as if it was him or her playing. Can you imagine watching your team’s game and being able to take the role of your favorite player?

If you combine the two recent events above, it is not crazy to think that in a not-too-far future enough it will be possible to personalize movies so that you can pick what the characters, heroes or villains, look like – something that has been a standard in the videogames industry for a long time now (think avatars). In movies, choosing whether you prefer the next Disney hero (or the next 007, or the next Joker) to be a tall, white male with dark hair, or an athletic Asian blonde woman, is no longer beyond imagination. A bit later in time, we could even think of using content-generating Artificial Intelligence (AI) to not only adapt a character to the looks of your choice, but even create your own story around it: some sort of build-your-own-movie (BYOM). Such customization has potential well beyond entertainment: brands already look to shape messages to key advertising audiences, and this could get to a new level thanks to generative AI.

Apart from the fun, we can reasonably expect that such personalization, or let’s better call it customization, will allow under-represented minorities -not just on the basis of race, but on any other basis- to feel represented and have role models of broader appeal. Just like in the earlier example of The Little Mermaid, every young kid could have their own Ariel, and every not-so-young-kid their own sports star or movie character. This might well be a major step in progressing towards a more diverse world, given everyone could feel represented by her favorite iconic character.

But… will it, really? I’d argue it will not. Giving the means for all to be represented is only half of the story.

In today’s form, entertainment can be a powerful driver for diversity to the extent that it broadcasts role models that are shared by all. Shared experiences are, after all, the basis of a group’s or a nation’s culture. However, ultra-customized entertainment (and more broadly, ultra-customized content) would have the opposite effect: to remove shared references. Had it been possible to tailor Ariel’s appearance in The Little Mermaid, one can easily imagine how we would have finished with tens, if not hundreds of different Ariels, all relevant to each viewer, but none relevant to a majority. It would have allowed everyone to feel represented, but none of those role models might have been widely accepted. Instead of favoring diversity, it would drive fragmentation.

A few months ago I discussed the concepts of normal and correct, and how the right course of action when something is correct, but not yet normal (in the sense of widespread) is to spread it. If we safely accept that diversity, or any other cultural value, is correct and not yet frequent enough, our efforts should go in the direction of spreading it. It is unclear though whether a hypothetical world of ultra-customized entertainment would achieve such target.

The Gartner hype cycle tells us that such high short-term expectations are likely exaggerated, and that the ‘Peak of inflated expectations’ is followed by a ‘Trough of disillusionment’. But even discounting such short-term hype, it is almost certain that, mid-term if not already short-term, ChatGPT will eventually transform the way we work and relate with machines in ways we don’t yet grasp.

As I mentioned in an earlier article, “technology has advanced exponentially […but…] humans don’t evolve at the same fast speed. […] As a result, over time the gap between the capabilities of technology, and those of humans who are expected to use it and benefit from it, widens”. Whether we become users of ChatGPT or another equivalent AI-powered product, it will be close to impossible for the average user to understand all the intricacies of the technology behind it. At best, he/she will become literate, maybe proficient in exploiting it.

Is that a problem? One could argue it is not. After all, every day we coexist and use machines that we know how to use even if we don’t have a clue how they’re made. Phones, cars, TVs, or even a simple light bulb are a technological mystery for the average user, despite the fact that we spend hours using them.

There is however a fundamental difference between using one of the physical items above, and using AI. Phones, cars, TVs or light bulbs deliver a function, but AI delivers knowledge. The average user of a functional product can easily judge whether it ‘does the job’, and will replace it if it does not. If your phone, car, or light bulb stops working or is no longer fit for purpose, you will change it for a new one. However, the average user (or again, consumer) of a knowledge-oriented product is less equipped to judge whether that knowledge is sound or not.

Over the last years, we have all witnessed how social media has facilitated the spread of misinformation. A recent study led by the University of Southern California showed that our ability or inability to discern accurate from inaccurate information is amplified by social networks and their reward-based incentives – for example the number of likes. Simply said, people may or may not be able to discern what’s accurate or inaccurate content when they read it on social media. But regardless, social networks help spread such content as they have created that habit of sharing in users, who seek external recognition via likes.

As the usage of AI-generated content becomes accessible to the broader public via user-friendly tools like ChatGPT and others, barriers to create content become much lower, it is harder to grade the legitimacy of that content, and the potential integration of AI tools with other downstream systems make it even easier to spread. If recent AI-generated pictures of Pope Francis in odd situations are rather inoffensive or even fun, the publication of an AI-generated picture in a Spanish newspaper showing two rival politicians in an improbable friendly attitude raised much greater controversy even if the picture was rightly footnoted as fake.

It is likely that soon enough we will no longer hear people say “I read it on Facebook”, but instead “ChatGPT said so”. In not too long there will be many other ChatGPT equivalents that generate content. What is the level of credibility we can grant to them? How to differentiate which can be considered an authoritative source of information, and which not? Individually and collectively as a society, AI-generated content will require us to raise our level of critical thinking.

Until now, we conferred authoritas based on factors like education and reputation. A reputed newspaper, or a frequently quoted academic would enjoy respect and credibility. It is much harder to evaluate AI-generated content: in the absence of explicit sources, the average user is more likely to rely on the brand behind the AI (e.g. OpenAI or Microsoft) than on verifiable credentials (e.g. the validity of the content that was used to train the AI model itself).

Further below [1] is a ‘conversation’ I had a few days ago with ChatGPT on this question, and where after a couple of probes, the chat suggested users should base their judgment on brand reputation, user experience, support and updates on the model, and general (undefined) accuracy and reliability guarantees. In my opinion, all these read like a commercial approach to the question, rather than an ethical one. ChatGPT places itself as delivering a function, rather than knowledge: the user is assimilated to a consumer.

OpenAI’s website refers to a January 2023 paper on “Forecasting potential misuses of language models for disinformation campaigns and how to reduce risk” which reflects on the risk and potential mitigating actions. Among them, it mentioned (although did not defend) potential government-imposed restrictions on data collection and access – something that has recently materialized in Italy., whose government mentioned that ChatGPT "exposes minors to absolutely unsuitable answers compared to their degree of development and awareness". A more recent paper from March 2023, on “Lessons learned on language model safety and misuse”, acknowledges that “when we first worked on GPT-3, we viewed it as an internal research artifact rather than a production system and were not as aggressive in filtering out toxic training data as we might have otherwise been. We have invested more in researching and removing such material for subsequent models”.

I am optimistic about the potential of AI tools like (for example) ChatGPT, but also remain mindful of the increased responsibility it requires, not only from the companies that built them, but also from each of us users. Among them, discerning the authoritas we should grant to AI tools, and the content they generate.

----

[1] Transcript of a conversation with GPT about how users could discern authoritative content from ChatGPT and other potential similar tools:

Question. Can you explain how you (ChatGPT) work, in simple terms?

Answer. Certainly! At a high level, I work by processing and analyzing natural language text input from users, and generating relevant responses in natural language that aim to answer their questions or provide assistance.

Behind the scenes, I was trained on a massive dataset of text using a deep learning technique called transformer neural networks. During training, I was exposed to a wide range of language patterns, allowing me to learn how to recognize and generate appropriate responses to different types of input.

When you interact with me, your text input is processed through a complex series of algorithms and computations, involving various layers of neural networks and statistical models. These models analyze your input and try to understand the meaning behind your words, as well as the context of your question or statement. Based on this analysis, I generate a response that I think best answers your question or provides the information you're looking for.

Of course, this is a simplified explanation of the many complex processes that occur behind the scenes. But overall, my goal is to provide a helpful and conversational experience for users by using advanced natural language processing techniques.

Q. Thanks. You say you were "trained on a massive dataset of text". What is that text? Is it the whole available content on the internet? Or something else?

A. That's a great question! The text dataset used to train me consists of a wide variety of sources, but it's not the entire content of the internet. Instead, my training data comes from a mix of different types of text, including books, news articles, academic papers, social media posts, and more.

The dataset is carefully curated to ensure that it represents a diverse range of topics, styles, and contexts, and that it is free from biased or inappropriate content. It's also constantly updated and refined to keep up with the latest developments in language and culture.

While the size and quality of the training dataset are important factors in my performance, it's also worth noting that my ability to generate helpful responses depends on the specific context and content of the input I receive from users. So, while I may have seen examples of similar questions or topics during my training, my responses are generated on a case-by-case basis and are tailored to the specific input I receive.

Q. Thanks again. How do you know your dataset is really "free from biased content"? I mean: that requires some judgment on what is or is not biased. And parts of that content could be plainly inaccurate.

A. That's a great point! While the dataset used to train me is carefully curated to avoid biased or inappropriate content, it's impossible to ensure that it's completely free from bias or inaccurate information. There may be cases where certain types of bias or inaccuracies are not immediately apparent, or where they only become apparent in certain contexts.

That's why my developers constantly monitor and analyze my performance, and use feedback from users to identify and address any biases or inaccuracies that may arise. They also use a variety of techniques to minimize the impact of bias, such as carefully selecting training examples, using diverse sources of data, and applying pre-processing techniques to remove or neutralize certain types of bias.

However, it's important to keep in mind that I'm a machine learning model and my responses are generated based on patterns and correlations in my training data. While I'm designed to be as accurate and helpful as possible, I may not always be perfect, and it's important to use critical thinking and verify information from multiple sources when making important decisions.

Q. Let's imagine that a few weeks or months from now there was another ChatGPT system available for use, built by a different team or a different company. How could I know which of you is better, and hence which of you I could recognize a higher authoritas?

A. That's a great question! There are several factors that could influence which version of ChatGPT is better for a particular use case, and it can depend on a variety of factors such as the quality and size of the training dataset, the specific algorithms and techniques used, and the ongoing development and improvement efforts of the team or company responsible for the system.

One way to evaluate the performance of different language models is to use standardized benchmarks, such as the SuperGLUE or GLUE benchmarks, which are widely used in the natural language processing community to compare the performance of different models on a range of tasks. These benchmarks evaluate the accuracy and performance of language models on tasks such as sentence completion, sentiment analysis, and natural language inference, among others.

Another way to evaluate the performance of different language models is to look at user reviews and feedback. Many companies and developers release chatbots and language models to the public, and users can provide feedback on the quality and usefulness of these systems. Additionally, some companies and organizations may conduct user studies or surveys to evaluate the performance and user experience of different language models.

Ultimately, the best way to determine which version of ChatGPT is better for a particular use case is to evaluate the performance and accuracy of each system on a specific task or set of tasks, and to compare the results to the user's specific needs and requirements.

 Q. I understand, but I'm a non-technical, regular user. I will struggle to compare the quality and size of the dataset (because I don't know the extent of each), don't know what a SuperGLUE or GLUE benchmark is, and user reviews are only a statistical but unqualified reflection of user feedback. How can I then know which of the several "ChatGPTs" I can grant higher authoritas?

A. You raise a good point! For non-technical users, it can be challenging to compare the quality and performance of different language models. In general, there are a few things you can look for that might indicate a higher quality or more reliable system:

1. Brand reputation: If a language model is developed by a well-known and reputable company, it's likely that they have invested significant resources in developing a high-quality system.

2. User experience: You can try out different language models and compare how they perform in terms of accuracy and helpfulness. If a particular system consistently provides more accurate and helpful responses, that could be a good indication that it's a higher quality system.

3. Support and updates: A company or team that provides regular updates and support for their language model may be more invested in ensuring its ongoing performance and reliability.

4. Accuracy and reliability guarantees: Some companies may offer guarantees around the accuracy and reliability of their language models, which can provide additional reassurance about their quality.

Ultimately, the best way to determine which language model to trust and use depends on your specific needs and requirements. You may need to try out different models and evaluate their performance on specific tasks or use cases to determine which one works best for you.

Q. Thanks!

A. You're welcome! If you have any more questions, feel free to ask.

In my experience leading teams in large organizations, I have found it helpful to complement company-wide values with an additional layer of ‘team tenets’, which guide the way in which I believe teams can be most effective and efficient, both internally and in their interactions with other parts of the organization. Given it has been well received in the past within my own and neighboring teams, I’m sharing them below, in case you find them helpful. Or as we say in Amazon, unless you know better ones!

1.      We drive actions and take decisions at the lowest appropriate level. Doing so at the highest possible level is inefficient and does not build judgment within the team. We escalate one-way door decisions, those for which there is lack of alignment, or those that belong above our level. For the rest, we take the shot and inform our manager if we need her to know or back us up.

2.      We manage upwards. Just as it is smart to be one step ahead of customer needs, it is smart to be one step ahead of management concerns. Managing upwards means communicating proactively and selectively, avoiding surprises, and proposing a solution for every problem we share or escalate. It does not mean being a ‘yes-sir’. Managing upwards trains our ‘Are Right, A Lot’, multiplies our ability to influence important decisions, and helps us be perceived as peers instead of reports. Management anxiety invariably multiplies our own workload: we manage upwards to reduce it.

3.      We prioritize based on value created, pain avoided, and team impact. We prioritize actions which create great value for customers or for the organization, which avoid great pain to the customer or the organization, which build the team’s skills, or which otherwise create a bottleneck for the team. We acknowledge that team development is always a time investment, and not a time expense. When challenging our own (or someone else’s) work, we do it to drive real improvements and on behalf of the customer, making efficient use of our scarce resources and time.

4.      We stretch our capabilities, not our hours. We build our capabilities –our professional muscle– through healthy ambition, effort, and practice. Growth starts with our own attitude: a strong attitude will help grow strong skills, but strong skills alone will rarely grow an attitude. While practice requires time, long hours are neither a synonym nor a necessary driver for professional growth.

5.      We take responsibility. We acknowledge that no other person in the company is tasked with doing our own job. What we do, will get done; what we do not do, will not get done. Because trying is not a synonym of doing, we choose to ‘do’, and do not settle at ‘try to do’. If we anticipate we will not be able to do our part, we humbly reach out for help. If we see a colleague struggling, we proactively offer help.

6.      We are frugal in the use of e-mail. Our work is to constantly improve CX, solve customer issues, and make the business sustainable. E-mail is not work: it is a means to achieve work. Wrongly used, e-mail is the #1 enemy of productivity. Hence, we are frugal in the use of e-mail: we include recipients that need to know (instead of nice to know); we rank them by distinctively using “To:” and “Cc:” (also in replies); we use relevant subjects; and we limit e-mail length to a maximum of one screen. We do not discuss by e-mail: we discuss by voice, and confirm by e-mail.

7.      We communicate for the receiver. An effective communication is measured by how much our counterpart understands and retains, not by how much we say. Just like we solve problems working backwards from the customer, we communicate working backwards from the receiver. We are well aware that our communication drives others’ first impression on our work. Hence our communication is crisp, clean, and tidy.

8.      We measure so that we can manage. Reporting is a means, not an end. It is a tool, not a solution. We need good reports, because we cannot manage what we cannot measure. However, we should not build reports for the personal comfort that ‘we did something’, and knowing that they will rarely drive any action.

According to the International Financial Reporting Standards (IFRS), an asset is “a present economic resource controlled by the entity as a result of past events”. In turn, an economic resource is “a right that has the potential to produce economic benefits”. And a liability is “a present obligation of the entity to transfer an economic resource as a result of past events”. In plain language, I like to think of an asset as something valuable that you own, and of a liability as something you owe or an obligation you are taking on going forward. If I asked the average person for examples of assets, aside from cash probably the most frequent responses would include a house, a car, jewelry, a or even some expensive goods like a TV or a smartphone. And indeed, you can typically extract value from them all: rent or sell the house, or sell any of the other items, and there you have cash to use for whatever else you want.

What the definition of asset does not capture sufficiently well is the obligations an asset carries: for example, the cost and effort of maintaining, or even more so increasing, their value. To ensure that a house keeps its value, you will have to maintain it: paint it every few years, fix the usual plumbing or electricity issues, or change or repair appliances when they break down. Otherwise, the valuable asset you thought you owns becomes an unattractive, highly depreciated bunch of bricks. Same reasoning applies to a car: not only does a new car lose a good 20% of its value as soon as it leaves the dealership, but owning it comes with a recurrent cost for insurance, maintenance, repairs, and gas. Even if you don’t use it, it’s likely that a big chunk of your maintenance costs would go into repairs (not using a mechanical system is a great way to ruin it!).

In addition to financial costs, there are personal and moral costs: time, workload, stress, or imposed obligations. In the simple cases above, if you own a house that requires frequent fixing, on top of the cost you will likely have to cope with the stress and burden of dealing with repairs. If you have an expensive car, you’re likely more reluctant to let it parked anywhere by fear of it being (in)advertently damaged by the driver parking next to it, and get more upset if that ever happens.

That is when assets cross the thin red line, and become liabilities.

The examples above are simple ones of personal life, but also exist in business. Imagine a company that builds a piece of software to solve a customer issue. After proper testing, the company launches that product, which quickly becomes a commercial (and financial) success. That sounds like an asset, right? But let’s imagine that the software is heavy to maintain, and difficult to integrate as the company keeps on launching other new products. Building it required 15 developers over 1 year, but maintaining it requires another five developers every year, and as the company grows, its architecture is so rigid that it will likely need to be rebuilt within 5 years. Is that still an asset? Not really: instead, it looks like technical debt, or as this article prefers to call it, technical liability: “sections of code, even entire systems, which because of their (poor) design are difficult to change, significantly hinder enhancement and are fertile ground for defeats”.

Same applies to public investment decisions. A recommendation adopted by the OECD council on March 12^th, 2014 already advised that “long-term operational and maintenance costs, which are often under-estimated, should also be fully assessed and planned for early in the investment cycle” – a big shout-out for administrations to estimate the Total Cost of Ownership before deciding on investing on an asset. Failure to do so will quickly turn any pretended asset into a money-draining liability. One frequently quoted example of such mis-estimation is the organization of the Olympic Games. One would naturally think that when a city decides to bid to become the host of the Olympic Games, it is because its authorities consider it an asset: an event that has value and will deliver long-term benefits. And yet, the economics of hosting the Olympic games are very unclear, with (for example) Montreal requiring 30 years to pay off the debt incurred from the organization of the 1976 games, or Sydney requiring $30 million a year to maintain its Olympic stadium.

We can often be blinded by what look like assets, but in reality are heavy liabilities that take a toll on our financials, time, or energy. If you don’t know it, I strongly recommend you to watch the 2007 movie “Concursante” (i.e. “The contestant”, currently available on Amazon Prime Video) which marvelously describes the story of Martin, a man who gets dragged into a personal hell as a consequence of winning the top prize of a TV show, valued in three million euros. Ninety minutes of asset-turns-liability story, worth keeping in mind.

This is a moment when many of us feel the strength, the urge, or simply the peer pressure to think about what we want to achieve in the next 365 days, and promise to ourselves we’ll do it. The bravest even promise it to others, as additional self-imposed pressure! My new year resolutions from past years have included the typical ‘lose a few kilos’, ‘get regular exercising’, or ‘play again the musical-instrument-which-I-used-to-play-when-younger-and-I-have-not-played-for-the-last-manymanymany-years’. And of course, I have succeeded in some, and frequently failed in others. Lack of time, lack of will, or simply a diminishing sense of strength as days pass and routine kicks in.

But… what makes a good new year resolution? Even more so: is there any such thing as a “good” and a “bad” new year resolution? In the last few days my wife, my brother and myself have chatted, and even teased each other, about our new year resolutions. And unsurprisingly, we have not gotten to an agreement of which are good vs. bad resolutions. Here are three different approaches: I’m keen to learn which one most resonates with you. And for those of you who know us three, you can even guess who advocates for each approach.

In Spanish, new year resolutions are called “propósitos de año nuevo”. And according to the dictionary of the Royal Academy of the Spanish language, the first meaning of the word “propósito” is “ánimo o intención de hacer algo” – an intent to do something. Accordingly, new year resolutions should somehow capture the activities or behaviors you intend to undertake in the new year. This fits quite well with the resolutions like ‘eat healthier’, ‘spend time with the people I love’, ‘travel’, and ‘take yoga or pilates lessons’. But… are these firm resolutions, or purely good intentions that risk of being short-lived? Quoting Jeff Bezos: “good intentions don’t work; mechanisms do”. And quoting Yoda: “Do. Or do not. There is no try”. I subscribe either of these quotes, hence you now know my opinion.

OK, so let’s take it one step further: rather than just an intention, every new year resolution might need to be formulated as a SMART goal (Specific, Measurable, Achievable, Relevant, and Time-Bound). By the end of the year, you should be able to say whether you achieved or missed each goal. This is closer to the second meaning of the Spanish word “propósito”: “Objetivo que se pretende conseguir” – the goal you aim to achieve. If you’re serious about your goals, you should probably include some associated elements like an action plan, resources, a routine, and some way of measuring progress. For example, if your goal is to complete a marathon within 2023 (a SMART goal), you want to make sure you set a training plan, buy yourself a good pair of running shoes and adequate gear, save time to train several times a week, and aim for longer distances every few weeks. If you need to do all that for every goal, it come become a tiring process in itself, right? Certainly yes… unless you find your strength in a larger purpose.

The sweet spot is likely there: new year resolutions that are not goals in themselves, but rather inputs to achieve a larger (life) goal. Coming back to the usual examples, you are more likely to achieve your new year resolutions to lose weight, exercise more, resume playing a musical instrument, or do volunteering work, if these feed larger life goals like (for example) to lead a lifestyle that allows to reach the age of 80 in good physical health; to enjoy a balanced life where you have sufficient personal leisure time; or to create new opportunities for less favored ones. Of course, that requires being somewhat clear about what your life goals are, but that’s a different debate altogether.

While the first of these three approaches could be criticized because of its laxity and lack of grip (overly emotional), the latter may sound like an awfully overengineered effort (overly rational). And how do you define your new year resolutions?

P.S.- For a different and very insightful approach to this same topic, I vividly recommend Giovanni’s post on his newsletter Beyond Tweets.

I recall a friend who would use the same mental framework in a more mundane context: to influence kids’ behavior. According to him, you should never ask kids what they want for lunch, or what they want to wear for school. Instead, you should ask them to choose between two meals you’d be OK for them to have, or two outfits you’d like them to wear. I don’t have kids of my own, but I’m interested in knowing whether this tactic works just in case I need to use it in the future!

Jokes aside, both examples show the advantages and risks of oversimplifying reality. Doing so can help you handle complexity (you may not want or simply can’t enter into lengthy logical debates with your customers or your kids), but it can also make you vulnerable to someone else’s conscious or unconscious biases: you may be forced to choose between a shortlist of options, when in reality there are many more. You’re just pushed into a false choice.

In a previous post I alluded at how today’s world is complex as it is busy and crowded with goods, experiences, and stimuli. Because we don’t like, or are not capable of handling complexity on a permanent basis, we tend to simplify reality to ourselves, for example via analogies, associations, and frameworks. That simplification can be perfectly OK, and help us make the right decisions by removing noise. But when simplification becomes oversimplification, comes the risk of manipulation. As Felipe González, former Spanish Prime minister, mentioned in a recent interview: “populism consists of giving simple answers to complex problems”. Here the link to the original article in Spanish and to a Google translation into English.

Think, for example, how many countries’ politics are still represented as a left vs. right choice, as if the whole spectrum of political debate was limited to one single topic for decision and two options. The reality is (fortunately) richer. For example, one’s choices on economic policy have nothing to do with one’s beliefs about -let’s say- social rights or cultural expressions. Whether one person is or not economically liberal is independent from whether that person is or not in favor of abortion, surrogate pregnancy, or (recurrent debate in Spain) the preservation of bullfights as part of the country’s cultural heritage. Political debate is, or should be, a multi-dimensional space. And yet way too often political debates seem to be reduced, or rather forced, into one single line where you can only move from left to right or vice-versa.

The same over-reduction to a single line exists when someone confounds liking with supporting everything anyone says, does, or stands for. Or disliking with hating and disapproving everything anyone says, does, or stands for. You may deeply enjoy the work of an artist and disagree with her values. Cheer for a given sports person and disapprove of things she does outside of sports. Support a given cause, but dislike with the way it is defended. Or vote for a candidate and disagree with some of the decisions she’ll make. This is because the universe of choice is, fortunately, not binary. Again, not a line, but a multi-dimensional space.

Way too often, debates in today’s world look like an accumulation of false choices trying to push us along a single dimension, like a train on a rail which can only go back and forth from one end to another, or (at best) stop somewhere in the middle. We are constantly pushed to take yes or no; for or against positions on any individual issue. And from that single choice, be forever classified in one of two sides, no matter what other future issues might come.

Don’t settle for a unidimensional society.

This campaign reminds us that environments, and the people who live in them, do not necessarily evolve at the same speed. In the last decades technology has advanced exponentially, and in doing so has driven very fast change in many aspects of society. Humans however don’t evolve at the same fast speed. As this Harvard Business Review article states, “computing power has increased more than a trillion-fold since the mid 1950’s, but our brains remain unchanged”. As a result, over time the gap between the capabilities of technology, and those of humans who are expected to use it and benefit from it, widens. Same applies for what I’d call ‘social complexity’. In an earlier post I defended that a higher number of interfaces and interactions drives higher complexity. From that perspective, today’s world is much more complex than that of a few decades ago. Commerce, means of transportation, media, or social networks have multiplied the possibilities for us to interact with our environment (each in its own way) and have created new needs, whether fundamental or superficial. By making our lives busier and more crowded with goods, experiences, and stimuli, we have made them more complex, and not all individuals adjust to that higher complexity in the same way.

I will not defend that complexity is good – if anything, the opposite. Actually, it is a responsibility of each and all of us to help reduce that complexity by removing or simplifying interfaces, and by turning technology more accessible. Indeed, technology can create complexity but it can also help reduce it. Think for example of smartphones and laptops: they are both equally sophisticated hardware devices, but smartphones enjoy a much more intuitive interface that makes them easier to use - just do the test with any 3-year-old! Today, voice activation aims at making tech devices (among which smartphones) easier to use by removing friction in the interaction between humans and machines.

But entropy tends to increase: despite our efforts, it is almost certain that over time the environment we live in will become more complex. In this context, I’d argue that it is necessary and healthy for us all to make an effort to evolve so that, over time, we are equipped to handle a higher degree of complexity.

Although I have so far referred to social or technological complexity, the same applies to a professional set-up. Companies and teams evolve. As they grow, they will likely become more complex. And if they grow very fast, complexity can grow fast, too. Re-organizations, processes, and (again) technology can help mitigate the speed at which professional complexity increases, but eventually individuals in the organization also have a responsibility to grow so that they can not only continue performing at their role, but be able to handle a growing degree of complexity. It is therefore a good temperature check to understand at which speed your professional environment evolves, and honestly assess whether you are evolving at a faster pace (congrats, you might be on a path to get promoted!), a similar pace (good that you’re keeping up), or a slower pace (in which case you may be at risk of falling behind).

Measuring your speed against the environment is equally relevant for businesses. Obviously, because if you move slower than competition, your business might be at risk. But more broadly, you can and should also measure your speed of change vs. that of other stakeholders, such as suppliers or customers. For example, if you are evolving at a pace that is drastically different vs. that of your suppliers (too slow or too fast), there’s a risk that your business goals will eventually misalign with those of your stakeholders, and hence will become less relevant to each other. Who needs to adjust to each other’s path, and how you need to support each other stakeholders to catch-up if one is falling behind, may depend on the specific context. Same applies to customers: great products have failed just because timing was not right, and the company had moved too fast (or too late) vs. its customer environment.

Are you, therefore, evolving at the same speed as your environment? And if you are falling behind, what actions do you need to take, or skills to build, to close the gap?