The question no course asks you

Sit down with a tutor for the first time and watch what happens in the opening minute. Before any teaching, before the material is even chosen, they ask some version of the same question:

What do you already know about this?

Then they listen, and often ask you to explain something back, and while you are talking they are building a rough map — what is solid, what is shaky, what you think you know and don’t, what you know but haven’t named. Everything they do for the next hour is chosen against that map.

Now count the courses you have taken that asked you anything at all.

It was never an oversight

The obvious reading is that course authors don’t care, which is unfair and also uninteresting. The real reason is more constraining: they had no way to use the answer.

A course is written before you arrive. It is a fixed artefact — recorded, edited, published — and it has to be, because writing it took months of an expert’s time and that cost can only be recovered by serving the same thing to thousands of people. Fixed content cannot respond to an answer. So asking the question would be theatre: you would type something thoughtful into a box and receive chapter one either way.

The one place the question does get asked is the prerequisites list, which asks it backwards. It does not ask what you know; it states what you should have, and leaves you to work out the difference. That is the same question inverted so that all the work lands on the person with the least information.

What the answer would actually change

It is worth being specific about this, because “personalised learning” has been a marketing phrase for long enough to have stopped meaning anything. When it has been used well, it usually meant adjusting the pace of fixed content, or which exercises you get. That is a thin version.

Knowing what someone already knows changes four things, and only the first is about skipping:

Where it starts. The obvious one. Not deleting material, but choosing the floor — which ideas can be used without being built first.

Which examples get used. This is the one that matters more than people expect. Explaining eventual consistency to someone who has operated a database replica is a different act from explaining it to someone who has not: for the first person there is a real event in their memory to point at, and it does more work than any invented scenario. A course that knows what you have done can reach for things you have actually seen. A course that doesn’t must invent a coffee shop.

What it borrows from. Existing knowledge is scaffolding for new knowledge, and the best comparison is always to something the reader already holds precisely. Someone who knows Git can be told what a merge conflict in a different system resembles, and be right. Without knowing that, a writer has to compare things to filing cabinets and hope.

What it leaves out. Not everything is worth covering for everyone, and which things are omissible depends entirely on where you are going and what you have.

Notice that three of those four are not about saving you time. They are about the material being different — better fitted, not merely shorter. The framing of personalisation as efficiency undersells it.

The evidence has been sitting there for forty years

In 1984 Benjamin Bloom published a paper reporting that students taught one-to- one, with mastery-based feedback, performed dramatically better than students in conventional classrooms — around two standard deviations better, which would move a median student to roughly the 98th percentile. He framed it explicitly as a problem rather than a finding: tutoring works, tutoring cannot be afforded at scale, so what group method could match it?

The exact magnitude has been argued about ever since, and later work generally finds smaller effects than the original two sigma. But the direction has never been in dispute, and neither has the reason it works. A tutor adapts continuously to one person. That is the whole mechanism.

So the constraint was never pedagogical. Nobody in education has spent forty years wondering whether fitted instruction is better. They have spent forty years unable to pay for it.

The part that is genuinely hard

I would rather not pretend the remaining problem is small, because it isn’t, and it is the part we find hardest.

Asking someone what they know is not the same as knowing what they know. Self-report is unreliable in both directions and the errors are not random. People overstate familiarity with things they have read about and never done — a subject you have absorbed through blog posts feels known in a way that evaporates on contact. And people badly understate things they use daily without ever having named them: ask an experienced engineer whether they know concurrency and many will say not really, while holding a working model most textbooks would struggle to convey.

And knowledge is not a list. It has structure — things you can do but not explain, things you can explain but not do, ideas held correctly and ideas held in a subtly broken form that will produce confident wrong answers later. A tutor finds the broken ones by listening to you talk, which is a very high-bandwidth diagnostic that nobody has convincingly replaced.

The best available signals are indirect: what you have actually built, what you have finished and how you did on it, what you were confused by and where. Those accumulate over time, which means a system like this should get meaningfully better at teaching you the longer you use it, and starts out only somewhat better than a good course.

That is roughly where we are. Educado asks about you before it builds anything — what you do, what you know well, what you have always meant to understand, how much time you have — and then generates a course for that person rather than for the average one. It is one question, and one question is not a tutor’s first hour. It is, however, one more than any course has ever asked you.

The thing that has changed is not that somebody finally thought of asking. It is that for the first time, the answer can be used.