The Machines Are Here. The Schools Are Training Your Kids to Be One.
Sixth in the series. So far we followed the money through the system: flat scores on doubled spending, a black box with no evidence, buyout chains, dollars leaving town, and your child's data leaving with them. This one looks forward. Because there is a final absurdity in the system we have not touched yet, and it is the biggest one: the schools are preparing children for jobs that will not exist.
Here are the receipts on the future of work. Not predictions from a newsletter. Numbers from the institutions that get quoted when governments plan:
- The World Economic Forum's Future of Jobs Report 2025: 39 percent of core skills will change by 2030. The fastest-growing skills: analytical thinking, resilience, creative thinking, AI literacy. The fastest-declining: manual dexterity, memory, precision. Read that list twice. The declining skills are exactly what a worksheet practices.
- McKinsey Global Institute: up to 30 percent of work activities could be automated by 2030. The jobs most resistant to automation are the ones requiring higher cognitive skills: problem-solving, reasoning, judgment.
- Goldman Sachs (2023): generative AI could automate roughly 18 percent of work globally, and the hit lands hardest on procedure-following roles. Data entry. Basic analysis. Document processing. The exact tasks that schoolwork has been drilling since the 1840s.
One sentence summarizes all three: AI automates procedures. It does not think.
Now walk into a classroom and look at the software the system spent $800 million a year on. Post 2 documented the product: adaptive multiple choice, cartoon characters reading prompts aloud, click the answer, next screen. Post 5 documented the machine listening to children read.
Do you see the problem? The ed-tech platform is a procedure-training system with a mascot. It takes a child and drills exactly the skills the WEF, McKinsey, and Goldman Sachs all say are being automated away. Follow steps. Match patterns. Select from choices. Optimize for the next screen.
The schools bought machines to teach children to be machines.
The part that should make you angry, and then make you think.
A child who masters procedural math is training to compete with AI. A child who loses to i-Ready's cartoon lessons is being prepared for nothing at all. Either way, the twelve years spent inside the system produced a skill set with a shrinking market price.
Meanwhile, the skills every planning institution says will matter are the ones no screen teaches: deciding which problem is worth solving. Choosing a strategy when no template fits. Checking whether an answer makes sense in the world, not just on the form. Explaining reasoning to another human. Connecting ideas across domains. Overseeing a machine and knowing when its output is wrong.
Those skills have a name in mathematics education: conceptual understanding. Understanding structure, not just steps. Building models, not just answers. Reasoning, not recalling. It is what the OECD calls transformative competency, and it is the thing AI cannot do for you, because directing AI is itself conceptual work.
The system knows this. The research has been saying it for decades. But procedural instruction is what scales, what standardizes, and what a platform can deliver at scale to 14 million screens. Conceptual teaching requires a knowledgeable human. So the system bought software instead of knowledge.
Follow the money and the choice explains itself: knowledge lives in teachers, and teachers cost salaries. Software lives on a license, and licenses cost renewals. The system optimized for the thing it could renew.
Here is the twist nobody markets: coding is not the escape hatch.
Every well-meaning adult says it. "Learn to code." Coding is procedure. It is the first thing generative AI is eating. Goldman's number includes it. The safe skill of 2015 is the automated skill of 2026. The answer was never a different procedure. The answer is the thing above procedures: deciding what to compute, evaluating what was computed, and knowing when the machine is confidently wrong.
That is not a technology skill. That is mathematics. Real mathematics, the kind where you build the model before you run the algorithm, and you check the answer before you trust it.
What we do about it
At Math Success, the entire pedagogy is built on the future-proof side of that line. Children build number lines before they use them. They draw bar models before they write equations. They reason about why an answer is true, not just whether the box turns green. The framework has five commitments, and every one of them maps to a skill the automation reports say is rising in value: taking students' ideas seriously, using multiple strategies and models, teaching conceptual before procedural, using the structure of mathematics, and embracing mistakes as information.
The system is training children for the machine economy's losers. We train children to be the ones who direct the machines.
One teacher with a stack of printable lessons and the knowledge to use them can start tomorrow. One parent can teach conceptual math at the kitchen table tonight. The system will not update in time. The people, updated with the right math, already can.
The machines are here. Do not raise your child to be one.
Do the math. Follow the money.
Sources
- World Economic Forum, Future of Jobs Report 2025 (39% of core skills changing by 2030; rising skills: analytical thinking, resilience, creative thinking, AI literacy; declining: manual dexterity, memory, precision)
- McKinsey Global Institute (up to 30% of work activities automatable by 2030; higher-cognitive roles least susceptible)
- Goldman Sachs, Generative AI report (2023) (~18% of global work automatable by generative AI; procedure-following roles disproportionately exposed)
- OECD Future of Education and Skills 2030 (transformative competencies framing)
- NAEP Long-Term Trend 2023 and i-Ready usage data (from posts 1-2; achievement flat, platform scale 14M students)
- WWC procedures context (post 2) and Curriculum Associates acquisitions (post 5) for the platform-behavior documentation
Every statistic above traces to a published report. Corrections with sources are welcome.