On Second Thought ■ Episode 2
In the 1990s, a computer science degree taught you C, operating systems, compiler theory, networking, and formal languages. You graduated understanding how a machine works from transistor to process. The tools were thin, the understanding was deep, and nobody asked whether you had a Kubernetes certification because Kubernetes did not exist and neither did the problem it solves.
In 2026, a computer science degree teaches you React. AWS has embedded its certifications into four-year university programmes at Western Governors University and Purdue University Global. The Kubernetes certification saw 250,000 enrolments last year, growing 49 per cent annually. The coding bootcamp market is worth $2.65 billion and projected to reach $14 billion by 2032. React dominates bootcamp curricula for the fourth consecutive year, with popularity up 116 per cent versus 2023.
Alan Kay said it in 2004: "Most undergraduate degrees in computer science these days are basically Java vocational training." Twenty-two years later, one merely needs to replace "Java" with "React." The observation has aged rather better than the curricula.
The Axiom
Learn the tool, learn the craft. The industry treats proficiency with a framework as evidence of engineering competence. A Kubernetes certification proves you can operate Kubernetes. It does not prove you understand what a container actually is: a process in a namespace with resource limits, running on a kernel that was doing isolation long before Docker wrote a CLI for it.
The certification market is worth $76 billion and growing. Cybersecurity certifications alone: $3.9 billion. The CKA (Certified Kubernetes Administrator) represents 54 per cent of Kubernetes-related job postings. The credential has become the proxy for competence. The proxy, however, measures familiarity with an interface, not understanding of a system.
The Origin
Universities once taught foundations because there were no tools to teach. When the only languages were C and Lisp, you had no choice but to understand memory, pointers, and recursion. The tools were thin. The understanding was mandatory. You could not hide behind an abstraction because there were not enough abstractions to hide behind.
Then the abstraction layers arrived. And with them, the vendors. AWS Academy operates in 6,800 institutions across 120 countries, with a $100 million commitment to education equity. Microsoft offers $100 per year in Azure credits per student. Google partners with 317 universities through Internet2. The curriculum follows the sponsorship. Not because universities are corrupt, but because funding shapes focus, and focus shapes graduates.
Operating systems, compiler theory, and networking have moved from required to elective in many programmes. Web development, cloud computing, and AI/ML have taken their place. The shift is not conspiratorial. It is economic. Graduates who know React get hired faster than graduates who know how a compiler works. The market rewards the tool, not the understanding.
One does not blame the universities. One merely notes that it is rather difficult to teach fundamentals when the funding comes from companies selling abstractions.
The Cost
MIT launched "The Missing Semester" in 2020. A course teaching shell tools, version control, text editors, and command-line automation. The course exists because MIT observed that its own students "lack knowledge of tools available to them" and "often perform repetitive tasks by hand." The standard CS curriculum, they wrote, "is missing critical topics about the computing ecosystem."
If MIT's students cannot use grep, one does wonder
about the rest.
The abstraction ceiling is real and well-documented. Developers who use Docker cannot explain namespaces and cgroups. Developers who deploy to Kubernetes cannot configure iptables. Developers who write React cannot explain what the browser does with their code after the build step. Docker, as one engineer put it, "feels like magic until your container gets OOMKilled or you cannot reach a port you swore was open. Then you realise you are not running a mini-virtual machine; you are just running a process in a very fancy cage."
When the abstraction breaks, and it always does, they have no layer beneath to fall back to. The foundation was never taught. It was skipped. And skipping foundations does not save time. It borrows it, at interest.
HackerRank's 2025 report found that only 22 per cent of developers are given time for learning and upskilling. 48 per cent must find the time themselves. Employers are "unsure early-career developers can code without heavy AI assistance." The abstraction ceiling is not just a technical problem. It is becoming a hiring problem.
The Question
Dijkstra wrote: "Computer science is no more about computers than astronomy is about telescopes." He meant that the discipline is about thought, not machinery. One does wonder what he would make of a curriculum built entirely around the telescopes.
He also wrote: "Universities should not be afraid of teaching radical novelties; on the contrary, it is their calling to welcome the opportunity to do so. Their willingness to do so is our main safeguard against dictatorships, be they of the proletariat, of the scientific establishment, or of the corporate elite."
One does note the phrase "corporate elite" with a certain quiet interest.
What if we taught the protocol before the framework? The syscall before the container? The language before the library? What if proficiency meant understanding what happens beneath the abstraction, not merely operating the abstraction itself?
250,000 Kubernetes certifications were issued last year. One does wonder how many of them could explain what a process is.
In the 1990s, CS taught foundations. In 2026, it teaches frameworks. Alan Kay warned us in 2004. MIT proved him right in 2020. AWS sponsors 6,800 institutions. Bootcamps are a $2.65B market. 250,000 Kubernetes certifications. The abstraction ceiling is real. When the abstraction breaks, and it always does, the foundation was never taught.