Foundations Are Still Load-Bearing
Why 'AI made the basics cheap' does not mean 'skip the basics.' The bottom of Bloom's changes role rather than disappearing — and cognitive agency is the reason the foundations still have to live in the student's head.
The Tempting Wrong Turn
If AI does remembering, understanding, and applying for free, the seductive move is to design courses that skip straight to the top — all analysis and creation, no memorization, no fundamentals. Offload the bottom of the pyramid to the machine and teach only the peak.
This is a trap, and it’s worth being precise about why, because the trap is subtle enough that thoughtful designers fall into it.
You cannot evaluate a claim you don’t understand. You cannot analyze a system whose parts you can’t name. You cannot create in a domain whose foundations you’ve never internalized. The higher rungs of Bloom’s don’t float free of the lower ones — they stand on them. A student who has outsourced all their factual knowledge to a model has no basis on which to judge whether the model is lying to them, and — per the last lesson — above the line, the model lies fluently.
So the foundations remain necessary. What changes is their role.
Same Knowledge, Different Job
Before AI, foundational knowledge was often the destination. The graded deliverable was the recall, the summary, the applied procedure. Getting the student to hold those facts and reproduce them was the point of the assignment.
After AI, foundational knowledge becomes the equipment — the thing that has to be in the student’s head not so they can hand it in, but so they can do the higher-order work that AI can’t. You still teach the anatomy, the syntax, the historical dates, the formula. But you no longer grade the recall of them as an endpoint. You grade what the student does with them on a rung where AI can’t substitute.
A useful test of the distinction:
Foundations are load-bearing when a student who lacks them cannot tell that the AI is wrong.
That’s the whole justification for still teaching the basics in one line. If the missing fact would let a student catch a fluent-but-false AI answer, that fact is equipment, and it has to live in their head — not in a browser tab they can consult, because the whole point is catching an error they weren’t told to look for.
The reflex argument for skipping foundations is “why memorize what you can look up?” But looking something up requires knowing that you need to — and knowing what the right answer roughly should be so you recognize a wrong one when you see it. A student with no internalized foundation doesn’t experience the moment of “wait, that can’t be right.” They accept the fluent falsehood, because nothing in their head contradicts it. Externalized knowledge can’t fire the alarm; only internalized knowledge can.
Cognitive Agency Is the Real Objective
Underneath all of this is the concept Kallas and the Estonia AI Leap designers keep returning to: cognitive agency. It’s the capacity to form your own view, test a claim, and take responsibility for a judgment — to be the author of your thinking rather than a conduit for someone (or something) else’s.
Foundations serve cognitive agency. Higher-order tasks exercise it. And “letting AI think instead of you” is precisely the loss of it — the student who accepts the model’s output has outsourced not just the labor but the authorship of the conclusion.
This reframes your whole design job. You are not designing courses to transmit information — AI transmits information better than any curriculum. You are designing courses to build and defend cognitive agency in a world that offers a frictionless way to give it up. Every foundation you teach, every task you set, should be justifiable by: does this make the student a more capable, more responsible author of their own judgments?
The Practical Rule
Put the two lessons of this module together into one design rule you can apply to any topic:
Teach the foundation. Assess above it.
Keep the fundamentals in your course — deliberately, unapologetically, because they’re the equipment for higher-order thinking. But move the graded evidence up the ladder, onto rungs where a student has to use those fundamentals to do something AI can’t do for them. Foundations in the teaching; agency in the assessment.
Module 3 turns to where this matters most and breaks worst: assessment.
A designer removes all factual content from an intro course, reasoning that 'AI knows the facts, so we'll only teach critical thinking.' Why does this backfire?
Reveal Answer & Explanation
Critical thinking has nothing to operate on — students can’t evaluate or catch AI errors in a domain whose foundations they never internalized, so ‘critical thinking’ collapses into deferring to the model.
Higher-order skills stand on lower-order knowledge; strip the base and the top has no footing. A student with no internalized facts can’t recognize when the AI’s fluent answer is false, which is the exact skill the course claimed to teach. The correct move isn’t to delete the foundations — it’s to keep teaching them as equipment while moving the graded work onto rungs (evaluate, create) where that equipment gets used. Teach the foundation; assess above it.
Up Next
Module 3 goes to the place the whole problem becomes concrete: assessment. Why the assignments you already have break under AI, and the repeatable pattern for fixing them.