IRRational Superiority: What We Talk About When We Don’t Understand AI
When semantic confusion hollows out tech discourse.
Sometimes I really start to wonder, when exactly did we stop using language to convey meaning and instead use it to create intricate screen savers in memory of a lost era when we used words for something we can’t really be sure of what exactly it was that we used it for.
AI Supremacy (
) says it is a blog positioned at the intersection of Artificial Intelligence, Technology and Business. That’s amazing if that was possible. For something to intersect, it has to be distinguishable and not be the same. Can yellow intersect with yellow? Semantically, no — because for something to intersect it requires semantically two things: we can meaningfully represent them in a chart and the line can cross. Yellow lacks these properties: A diagram with y-bar: colour and x-bar: yellow is a point and not a line. That’s why yellow cannot intersect with red at orange — that is semantically a contradiction. You need two things that we express using charts and these two things have to be different or in contrast. A demand curve and a supply curve intersect. But it makes no sense to say cars and BMW intersect. How would that be possible? Something cannot intersect with itself. So unless AI is not technology but not-technology (i.e. a natural phenomenon), it cannot intersect with technology.In this sense, the discussion about AI often mirrors financial derivatives at the edge of the 2008 crisis: everything references everything else, and nothing touches ground. I am not saying Michael caused this problem or has any bad intentions. His latest article simply exemplifies how far-reaching this problem is.
One more comment:
. I am not prescribing rules on how you or anyone should name a publication. That’s your call but some feedback should he allowed. Perhaps starting with a question.What supremacy?
It is a superlative of higher or above, to say higher than high (which is in itself already a problematic wording) or the highest (that kinda works). It denotes the position of somebody or something occupying the highest position in a hierarchical order. AI supremacy implies that in some hierarchy AI is the one on top that controls the elements in this hierarchy unbounded and unlimited in its capacity to rule. It is not the same as sovereign. Our political theory, for instance, conceptualizes government as sovereign but not supreme. Supreme is hierarchy — i.e., one is above, that means others are below. Supremacy requires normative judgement of better and worse, powerful and powerless. A sovereign is also a superlative like supreme but binds this concept to a person as opposed to a defining quality of who or what one is.
There is thus a difference in saying "the king is supreme" or "King such and such is supreme." That is why the nuance of calling government sovereign or supreme matters. Attributing a person with supreme privileges — even if it sounds contradictory — is limited because a person is limited. This way it is highly logical. Neither AI sovereignty nor AI supremacy are ontologically coherent expressions. It would require some hierarchy, whether political or simply a system of categories, that has a normative judgment placing AI at the highest or best position and AI — either specific AI (King Henry so and so) or AI by nature (kingship) — holds this privilege.
And neither makes any sense.
Both terms derive from Latin, but they diverge significantly in meaning and function because of their respective historical contexts, especially in political philosophy:
Supremacy comes from superemus (highest, above all). It denotes absolute dominance in a vertical hierarchy — it's a status.
Sovereignty comes from superanus (above), but via medieval political theory, it evolved into a legal and moral concept of authority over a bounded domain. It implies responsibility, not just the position of principal without this.
In the article, he refers to a slide from a keynote address given at an investment management conference: "Coatue View on the State of the Markets." One of the cofounders describes this company Coatue as follows:
“I founded Coatue in 1999 with the vision of building a leading investment firm focused on lifecycle technology investing across both public and private markets.”
Lifecycle technology? What is that?
“Lifecycle technology” is a non-concept. It sounds like it wants to gesture toward technology across the business lifecycle — i.e., investing in tech companies at different growth stages (early, growth, late stage) — but that’s a financial strategy, not a category of technology.
Saying you invest in “lifecycle technology” is like saying you invest in “duration assets” or “cyclical cycles.” It fills space without adding meaning. It pretends there's a class of tech defined by its temporal trajectory, which is a confusion of object (technology) and valuation strategy (when you invest) or investing in technological technology (because a technology application also has a lifecycle).
The graphic concerns 40 companies called the Coatue Fantastic 40 — apparently a list, a projection of the top 40 of the corporations in 2030, also indicating IRR (Internal Rate of Return).
The author says:
“Internal Rate of Return (IRR) is a financial metric that calculates the discount rate at which the net present value (NPV) of all cash flows from a project or investment equals zero. That is, it represents the effective annual rate of return on an investment, considering the time value of money.”
That’s a bit confused but I don’t blame him. The reason is invoking concepts that are not directly what this is about. For example, a speculative overview is by definition not an effective rate of anything. It is the opposite of an effective rate of return.
IRR implies projected returns from actual capital deployment, assuming reinvestment and cash flow timing.
CAGR (compound annual growth rate) simply tells you the average rate of growth between two points in time. It's agnostic to mechanism, liquidity, or risk.
Calling it Implied CAGR would at least admit: this is speculative, using a method to describe historical price movement. But IRR suggests something deliberate, replicable, and financeable — which, since the list includes ETH and BTC, now makes a second error. ETH and BTC do not signify holding property rights or other rights. In itself it has no intrinsic value. Applying an IRR to an instrument that doesn’t meet the logical condition of IRR is highly questionable. The math may work, but it has nothing to do with a method called IRR, which is not applicable to such assets. If they say, "we annualised and discounted some speculative returns," then it would be a correct description of what they did. They did not calculate IRR for BTC because that is impossible.
The phrase “Implied IRR” masks a speculative CAGR and dresses it in the appearance of discounted finance. It implies rigor where there is none, and it falsely equates cash-generative firms with price-volatile tokens. It is a misuse of financial terminology that blurs speculative intuition with structured valuation.
This becomes very problematic if you look at the companies they have chosen:
Uber is a platform-mediated market, not a technology product.
Spotify, Netflix, Shopify — their stack is cloud infrastructure (bought from AWS, GCP), commodity recommendation engines, UX convenience. They are packaging companies that also trade on network effects, not tech.
The word "tech" is hollow unless anchored in structural causality:
Is tech the cause of the value?
Is the tech proprietary, infrastructural, and modular?
Can the effect be traced back to a technical system that could be independently assessed or replaced?
If not, calling something "tech" is just:
Branding
Investor seduction
Or a lazy abstraction for platform effects
Uber
It runs on a platform created by tech — but not an open protocol, not interoperable, no reproducible infrastructure. If it used technology as a value proposition, we would have 100s of competing Ubers. The fact we don’t means the tech is not what creates value. It is exploiting human behaviour when technology doesn’t allow for substitutability and open access. A mobile device sold as no-brand is tech; an iPhone is the same tech but doesn’t compete as tech. Hence why it costs more.
Network effects stem from coordination and trust problem between driver and customer who don’t know each other and could not set up a contract to do business. They let you find someone adhering to their protocol. They control through constraining our free choice to take it or leave it. That is not bad as such - neither good or bad. That comes later.
So: it's not “tech” in the sense of an defined, portable stack. It's a closed behavior-routing company.
And if we actually could build driverless cars, their value proposition collapses. Then constraint becomes fleet management and they can’t offer value the same way as today.
Revolut
Convenience built on existing rails (Visa, Mastercard, SWIFT).
UX layer.
Not technically independent or irreplaceable. Not "tech" in a foundational sense.
Salesforce
SaaS (outsourcing), but more of a CRM aggregator than a technical innovation engine.
Binds clients via data captivity, not standards or protocols.
Less “tech,” more enterprise Stockholm Syndrome.
OpenAI
Still coherent, for now: model, inference, and interface are vertically integrated.
BTC or ETH
Not a financial instrument and it entitles to nothing — this is where the table collapses into incoherence. It cannot be valued like financial instruments.
Nvidia
Actual tech company: Chip design + manufacturing + tooling.
Standard for others to build on.
Since their list of companies hardly includes any real tech and mostly consumer UI experience tools, we really need to ask a question: can we secure our future this way? No real advanced technology breakthrough can be expected if our markets produce consumer brands dressed up as tech when they are not. It is a risk to these companies as well.
Just a thought. It is not so easy to fix because it starts with clear terms in describing what is and what isn’t. That should be easy and we take it for granted — but it is not so.
#myndOS #shemot