Rex Yang

Some notes addressing a variety of things I’ve seen and experienced firsthand, and a reminder of a handful of things to be very careful about. This is neither a formal essay about my reasoning behind a list of strict requirements, nor encouraging taking any of the following directly as advice.

Let’s start with navigating purpose.

“Purpose” is among the most grounded, widely discussed, but easily confused ideas in modern life. You especially do not want to misunderstand it in entrepreneurship, because once you do, everyone involved begins paying for the confusion. One way to see the problem is to notice how many failures are better explained by purpose than by skill, and how “purpose matters more than skills” survives mostly as a motivational slogan while being practiced far less often in reality. Some of this has to do with people making excuses for weak self-accountability, prestige-whoring, and being distracted all the time. But a less discussed problem is how hard it can be for certain people, including many capable ones, to distinguish purpose from the systems that happen to reward them; they think it ‘takes them lots of time/money’ and other things! I became increasingly curious about how all this confusion was going on, comparing my observations while in the US from 2002-2011, China from 2011 to 2021, and then the US again from 2021 till now.

Some groups appear especially vulnerable to this confusion. They try so hard to do three things: 1. Constantly stressing about their own lack of purpose but with minimal forward-looking intent; 2. Relying on external people/systems, preferably the most credible ones available to them, to ‘help them discover’ what they should do; and then 3. Mistaking that externally assigned direction for the actual reason they are doing it, only to blame the system once the rewards stop arriving at their expectation. Some third-culture IELTS tutor probably has a story or two about this. You can observe versions of this at nearly every economic class and stage of life. In some sense, it is related to the degree to which a society organizes important functions around unified, top-down, easily legible metrics, which easily become counterfeit purpose for the average individual.

• Exhibit 1. While a fair reason for Gen Zs to go study abroad is that a highly centralized education system with a large systemic obsession with grades and rankings was too painful for them to endure¹, little do they realize the metric-obsessive mentality which they themselves possess will bother them all along, whether they admit it or not. The discussion of university rankings lasting late into our 30s has shown how hard it is to change this mentality. This is no different for the older generations who wonder why children they’ve pushed so aggressively toward measurable success seem strangely passive or lost afterward, and further borne out in recent years, when today’s graduates from even highly ranked universities encounter a labor market being rearranged by technology.
• Exhibit 2. American universities have seen arguments over grade curves, publication pressure, and administrative incentives, which can still look mild compared with institutions elsewhere with minimum graduation or pass-rate expectations effectively imposed from above, and instructors are left to to ‘figure it out’, sometimes through practices that would plainly be called institutional fraud.²
• Exhibit 3. Some companies, including certain automakers, mass-produce consumer products largely to overtake competitors across a respectable-looking list of metrics, but end up dismissed by the market as sacrificing reliability and basic safety to do so, because the metrics replace the purpose rather than serve it; meanwhile, the ThrustSSC being the fastest ground vehicle on earth is successful precisely because its narrow metric followed from its purpose.

Metrics matter in many senses, and a metric-obsessive mentality is not necessarily fatal; for people with genuinely critical judgment, its pros and cons can often be navigated, and it will be interesting to see how far things develop. But a serious concern may be how much we underestimate its growth in Silicon Valley today, until we repeatedly encounter the kind of opportunistic founders who infer the size of an opportunity simply from the popularity of a topic—say, the number of talks, research proposals, or posts about it on X—yet have almost nothing to say about something that appears important and technically interesting simply because the visible market looks small! My stance is that you neither want metrics to be the priority consideration for a serious purpose, nor the sole factor dictating purpose, especially when starting companies.

Opportunists³

Being opportunistic on certain occasions, I never saw the real issue to be Artificial Intelligence creating a large group of opportunists, but instead we should really be concerned about how this era is creating the worst opportunist types ever: Opportunists who don’t admit it, and opportunists who trick both others and themselves. It will only get worse as lack of self-realization and an obsession with metrics-based judgement grow among young people.

Metrics and Judgement

Without a clear purpose, judgment loses its standard, so people retreat to metrics as substitutes for knowing what actually matters. Some know to what degree they should judge based on indicators/metrics. But it is crucial to understand that data sets by their very nature are backward-looking, and creativity by its very nature is forward-looking. An illustration of this is how observers who judge on failed company cases think they’ve mastered how NOT to run a company, but only find out how different realities are when they become the operators building one themselves, especially if they were given the same idea or problem. The distinction is not whether operators have more information than observers, but operators are forced to recover first principles and define what success means under real dynamic constraints, while observers can avoid that responsibility and judge through whatever indicators are easiest to see. We are all, more or less, armchair experts in various periods of our lives; while we are constantly learning to move beyond that, choosing to be a ‘supposedly smart’ but backward-looking person in a forward-looking industry can be one of the worst life choices a person can make.

Metrics can be very misleading, along with the growing noise on the internet. Today’s generation tends to base their judgment on metrics more than they realize, thinking metrics can lead to more certainty but only to cause more noise. In a niche sense, you might not want to use startdowns.ai to determine real company potential, especially during broad hostility toward “AI slop,”⁴ since startups systematically getting more downvotes may cause a high likelihood of some good companies being misinterpreted or mistaken. See how the uncertainties magnify when it comes to many companies being built supposedly based on nonconsensus truths; however, they can be solved or seriously examined by operators through deep learning and sustained focus.

Clarity of purpose is what makes deep study and sustained focus effective enough to turn an outsider into an insider⁵, or a generalist into a better specialist and, eventually a better generalist again. The current obsession with LLMs, however, has made it unusually easy to move across many subjects quickly enough to feel informed without remaining anywhere long enough to become so. That makes this a useful moment for self-examination, no matter how well we think we understand the models themselves. The broader AI panic may ultimately make this point more convincingly than any call-to-action from a neurodivergent AuDHD person who has gone through several lessons.

Unlike LLMs, which can process certain categories of work in parallel with little apparent cost, humans eventually reach a point where multitasking produces diminishing and then negative returns. Beyond that threshold, breadth begins to work against insider knowledge, original judgment, and real innovation; when generalists without specialist depth repeatedly discover ideas that have existed for decades and briefly mistake rediscovery for invention, the result is not evidence that we have become unusually creative, but shows how shallow generalization can produce a dangerous feeling of completion before we have stayed with a problem long enough to discover any secrets.

The AI Panic in the job market

By now we have accumulated plenty of reasons to doubt the simple story that LLMs are replacing humans wholesale, while many recent layoffs appear at least as connected to cost structures and policy adjustments as to technological displacement itself. Yet this has done little to reverse a mentality among young college kids that the safest response to uncertainty is to multitask and grasp as many opportunities as possible. The irony here is that this supposed risk reduction can make people more replaceable, while an insider who stays deeply focused on one problem at a time will usually discover far more work that AI cannot easily perform, either because it requires judgment, context, taste, trust, physical interaction… things that only become visible after prolonged exposure. From there, the choice is fairly open: exploit the human advantage directly, help build better applications of AI around what you have learned, join a mature company where that depth matters, or start something with the right people.

Now let us return to the purpose question: Why start a company? Two common answers today are that “now is the best time to start”—usually some mixture of technological optimism and the ‘fear of missing out’ (FOMO)—and that building a company is one of the fastest ways to do something impressive enough to “differentiate yourself from everyone else”. Neither answer is necessarily wrong under the current technological landscape; both can even be rational. But what have we done when they stop being secondary considerations and quietly become the reason?

We are living through an unusually favorable period for starting companies, but perhaps also an unusually favorable period for wanting to be a founder without wanting anything specific badly enough to justify founding one: The prestige hierarchy that once concentrated around universities, consulting firms, banks, and large tech companies has long migrated into entrepreneurship and has seen considerable growth under the influence of the advancements of Artificial Intelligence. This helps explain several strange tendencies in the current startup ecosystem.

• A large number of companies now fall into the “shovel seller” category, to the point that shovel-selling has formed an over-invested loop of its own, reaching ever higher valuations while solving remarkably little outside the ecosystem that created the demand for those shovels in the first place.
• Automation has become too easy, making it increasingly common to confuse “this can now be automated” with “this deserves to become a company.” And with that ease comes one of the lazier confusions of the current startup cycle: that because a workflow can be automated, the workflow itself must have been the problem. Several industries suffer deeper structural problems that cannot be mitigated through workflow automation. Take real estate rental as an example, where stakeholders’ responsibilities and systems were never designed to share state or coordinate cleanly through the digital infrastructure now sitting on top of them, and thus automating one visible step may save several minutes while leaving the actual coordination failure on the user end almost untouched.⁶
• One increasingly common form of prestige-whoring is the founder wannabe who says he wants to start a company primarily because he wants to “do his own thing,” “work for himself,” or become independent, which is almost delusional. The desire for independence without an appetite for responsibility is not entrepreneurial independence at all, but the fantasy of authority without obligation, made especially attractive to a generation that has grown up too comfortable and told for years that personal freedom means minimizing the number of things one owes to other people.
• Since technology has become so capable, it is now fashionable to insist that startups begin with the smallest team possible, because small teams supposedly move faster—as they say. Perhaps; but a market full of tiny, fragmented teams also happens to be ideal for the giants, which can let them absorb risk, test markets, and build around their platforms before buying, copying, hiring, or crushing whatever proves useful. Eventually you get tens of thousands of small teams that look like decentralization from below and consolidation from above: The waves of startups being erased by each major OpenAI release are less a demonstration that startups are doomed than a reminder that speed is not defensibility, and technological dependency is not technological depth. While many young founders appear to know this in theory, they continue to build companies that assume the opposite in practice.
• Technical prestige whoring is how a large part of today’s tech founders constantly chase whatever appears technically impressive, until they discover halfway through that conventional business experience has taught them little about how capital, incentives, institutions, and public adoption actually interact; this is one way of showing some of today’s young tech founders not learning & practicing enough economics, although they may seem forward-looking. Young tech founders also overlook the fact that the more widely available a capability becomes, the less distinction comes from merely possessing it: At this very moment in history, AI chatbots aren’t inherently interesting since Q3 2025. The same thing is happening to AI agents now.

As these ego-empowering cycles grow, it also happens to work remarkably well for the tech giants: opportunists enter startups during a boom, build on top of infrastructure owned by increasingly powerful companies, and, once their own projects go bust, often return as better-credentialed employees. No elaborate conspiracy is necessary, and forget customer feedback loops—this may be one of the most interesting loops (or divide & conquer rigs, I would call) Gen Z gets to watch operate and contribute to in real time!

While it is easy to blame tech giants, investors, accelerators, or startup culture for this, it is probably more useful to reconsider our own definition of purpose, or simply why we do things at all.

We have heard often enough that purpose should not be treated as an object hidden somewhere inside the self, waiting to be discovered through sufficient introspection. The suggestion earlier about taking a position, attaching to something concrete, and remaining there long enough for the shallow reasons to wear off may also turn out to be a more enjoyable experience, similar to how people into trading might turn out to be less miserable by autosuggesting they are purely passionate about it, instead of spending ten years treating it as a temporary instrument for becoming rich enough to finally start the company that will supposedly make him independent.

Having the technical foundation is obviously not the same as having permission from reality, which is why outlier technologies can so easily become inflated purposes: we can talk endlessly about the next engineering breakthrough, only to discover that ordinary human prejudice have no particular obligation to keep up. Some technologies are difficult because the physics is hard; others because people have already decided what should look normal, and are perfectly happy to call everything else unrealistic. For some entrepreneurs, altering prejudice may therefore be a serious purpose in itself, although there is probably little point in lecturing everyone into becoming more open-minded; one upside of being a responsible founder is that we can build defensible companies that attract attention, change incentives, and make previously unrealistic technologies economically and politically possible, perhaps long before the people who dismissed them are willing to admit they were wrong.

But only after these purpose questions are reasonably clear does it make sense to talk seriously about paths, leverage, and what we actually want or how we want to impact the lives of others. This reminds us of the old inscription Cliff Stoll once recalled: http://www.donath.org/Quotes/AllTruthIsOne/