There is a popular game in finance where people arrange their profession into a pyramid: private equity at the apex, then public-market fund managers, banks, venture capital, and investment banking at the base. The higher you sit, the more you combine independence over capital, depth of knowledge, and ownership of outcomes.
It is tempting to build the same pyramid for computer science. Deep Tech and frontier research at the top, because it creates the capabilities everyone else consumes; then product technology that reaches the masses; then the enterprise software that is essential to the corporate world; then the IT services that work on a massive scale. The hot roles slot in neatly: the AI engineer industrialises, the data scientist straddles the middle, the application developer anchors the base.
Read more: Is private equity really the pinnacle of finance?
The pyramid is a good hook. It is also wrong, and the way it is wrong is the story.
The pyramid hides a second axis
Take two people. One is a mid-level researcher inside a frontier AI lab. The other is a brilliant software engineer. The pyramid says the first ranks higher, sitting closer to the apex of capability. But the engineer may be more skilled, more irreplaceable, more in command of their own outcomes. What the researcher has is not superior ability. It is proximity to scarce inputs: frontier compute, model weights, proprietary data, and the people who understand them.
So which one is higher? The question is malformed, because two different things are being measured and the pyramid collapses them into one. There is the platform you sit on, meaning whether the institution around you controls scarce inputs. And there is your individual scarcity, meaning how replaceable you are inside it. These vary independently. A pyramid forces them onto a single rung; reality spreads them across a plane.
This invites a fair objection. If a thousand engineers at a lab all hold the same access, access cannot be what separates them from each other. Too many people have it. True enough. Access is scarce between populations, not within them. A lab might have a thousand people with frontier access; the global pool of capable engineers is in the millions. So access is brutally scarce across the field and nearly worthless as a tiebreaker inside the building. What rescues the individual is that proximity is an option on becoming scarce: sitting next to frontier work, you absorb what is never published and build intuition that does not transfer. Treat the seat as a salary and you keep a comfortable floor. Treat it as a learning position and you convert platform into scarcity.
