Your practice is lying to you: the “blocked problems” scam (and how AI can fix it)

If you do 30 identical problems in a row, you will feel smart.

You will also be training your brain to get good at one thing: following a groove you just carved.

Coaching classes love this groove. It looks like productivity. It is easy to timetable. It keeps parents calm. It produces that satisfying feeling of momentum.

And it quietly creates a student who collapses the moment the exam shuffles the questions.

This is not motivation talk. This is cognitive science.

Why blocked practice feels so good

Blocked practice means you repeat the same type of problem again and again. Same chapter. Same formula. Same steps. Only the numbers change.

The first few problems are slow. Then something happens: you start flying.

That “speed boost” is not mastery. It is short-term pattern matching.

When the problem type stays constant, your working memory holds the procedure. You stop searching. You stop choosing. You stop noticing details that actually matter.

So you get a dangerously convincing signal: “I know this.”

The research word for this mismatch is fluency. Fluent performance during practice is not the same as long-term learning.

Cognitive scientists call this the illusion of competence. Practice that feels easy can create confidence that is not earned.

This trap is old. The industry around it is new.

The exam does not come in chapters

Real exams do something simple and brutal: they mix.

JEE does not give you a clean page labeled “Projectile Motion: 20 questions.” It gives you a sequence designed to force selection: Which concept? Which approach? Which approximation? Which formula is even relevant?

That selection step is the entire game.

Blocked practice removes that step. It makes you look good in practice and dumb in the exam.

So the correct question is not “How many problems did you do today?”

The correct question is: How often did you have to decide what to do?

Interleaving: the practice method coaching avoids

The opposite of blocking is interleaving.

Interleaving means you mix problem types so your brain must identify what the problem is asking before it can solve it.

Example:

• One kinematics problem
• One NLM problem
• One work-energy problem
• Then a kinematics problem again, but not the same flavor

This feels worse.

Your speed drops. You make “stupid mistakes.” You feel like you are forgetting.

That discomfort is the point.

Interleaving forces discrimination. It builds the mental skill exams demand: choosing the right tool under uncertainty.

And this is not a motivational claim. We have data.

In an influential study, Rohrer and Taylor showed that shuffling different types of math problems improved later test performance compared to practicing one type at a time (Rohrer & Taylor, 2007).

Kornell and Bjork found a similar effect when learners studied different categories of paintings. The mixed condition felt harder, but the test performance was better (Kornell & Bjork, 2008).

The pattern shows up across domains: when the test requires discrimination, practice that trains discrimination wins.

Spacing: your brain needs time to forget

Now add another insult: spacing.

Spacing means you revisit a topic after time has passed, instead of finishing it and “closing the chapter.”

This also feels worse.

When you come back after a few days, you are not warm. The steps are not in your RAM. You have to rebuild.

Again, that struggle is not a bug. It is the mechanism.

A large meta-analysis found distributed practice (spaced repetition) reliably beats massed practice across many verbal learning tasks (Cepeda et al., 2006).

Spacing works because it forces retrieval after partial forgetting. That retrieval strengthens memory far more than re-reading or re-watching.

Retrieval practice: the uncomfortable superpower

Most students confuse “exposure” with “learning.”

They watch solutions. They read notes. They nod.

It feels like progress because your brain recognizes the material. Recognition is cheap.

What you need is retrieval: pulling the idea out of your head when it is not in front of you.

Roediger and Karpicke demonstrated that repeated testing improves long-term retention more than repeated studying, even though students often predict the opposite (Roediger & Karpicke, 2006).

Dunlosky and colleagues reviewed major learning techniques and found strong evidence for practice testing and distributed practice, while many popular habits had weak evidence (Dunlosky et al., 2013).

So if you want a ruthless summary of the science:

• If it feels smooth, it might be fake.
• If it forces retrieval, it is probably real.

Why coaching systems fight these methods

If interleaving, spacing, and retrieval are so good, why are they not the default?

Because they are inconvenient for factories.

Coaching systems are built around:

• linear schedules
• chapter completion
• uniform homework
• batch-wise “coverage”

Interleaving breaks the idea of “we finished this chapter.”

Spacing breaks the timetable.

Retrieval practice exposes gaps in public. That is emotionally messy inside a classroom that wants to look efficient.

So the system optimizes for something else: visible activity.

It optimizes for attendance and compliance.

The student pays the price later.

The student experience: why this feels like failure

Here is the cruel part.

Interleaving and spacing make you feel like you are doing worse precisely when you are doing better.

That can create panic, especially for JEE/NEET students living inside rank anxiety.

So a humane learning system has to do two things at once:

1. push the student into productive difficulty
2. protect the student from despair while it happens

Coaching classes usually fail at both.

They either keep things “easy” (so nothing sticks), or they make it hard without feedback (so students break).

This is where the right kind of AI tutor can be genuinely transformative.

What AI should do (and what it should never do)

Most “AI education” today is a solution machine.

You type a question.

It gives you the answer and a pretty explanation.

This is candy. Sometimes you need candy. But if you live on it, you rot.

A learning AI should behave less like a teacher who performs, and more like a coach who designs practice.

That means:

1) Force the choice before the solution

Before giving steps, the tutor should ask:

• What chapter do you think this is?
• Which principle applies?
• What is the one unknown you must express?

This is not annoyance. It is training.

2) Interleave on purpose

Instead of “today we do capacitors,” the tutor should build sets that mix:

• capacitors
• current electricity
• electrostatics

Not random mixing. Targeted mixing.

When you miss a problem, it should not just give the answer. It should update the mix.

3) Space based on forgetting, not calendars

A good system does not say “revise every Sunday.”

It measures what you can retrieve. Then it schedules the next encounter before the memory collapses, but after some forgetting has begun.

This is the difference between revision that feels comfortable and revision that actually prevents forgetting.

4) Give hints that preserve struggle

There is a skill to helping.

If the hint reveals the key step, the student stops thinking.

Better hints are constraints and questions:

• “What assumption are you making about friction?”
• “Can you write energy conservation before computing numbers?”
• “What happens if you take the limit as time goes to zero?”

The hint should keep the student in control.

5) Measure learning, not time

A coaching app loves “minutes studied.”

A real learning system loves “successful retrieval after delay.”

Those are different worlds.

A concrete 14-day experiment for JEE/NEET students

If you want proof without trusting anyone, run this experiment.

Choose one chapter you think you are “strong” in.

For 14 days:

• Do short mixed sets (10 to 15 questions) instead of long same-type marathons.
• Include questions from 3 chapters per set.
• Do not watch solutions unless you have tried for 10 minutes.
• The next day, start with 5 questions from the previous day’s set.
• Every fourth day, do a mini-test: 25 mixed questions, timed.

Track only two numbers:

• accuracy on the mini-test
• time per question on the mini-test

Ignore how you feel during practice. Your feelings are the unreliable narrator.

If you do this honestly, you will notice something uncomfortable: your practice sessions will feel slower, but your mini-test performance will climb.

That climb is what you actually want.

The bigger point: learning is a design problem

Students blame themselves too quickly.

“I’m not disciplined.”

“I’m not consistent.”

“I’m not smart.”

Sometimes the student is tired. Sometimes the student is scared. Sometimes the student needs help.

But often the student is simply trapped in a bad practice design that produces a false feeling of competence.

Blocked practice makes you feel like you are winning. It is often a mirage.

Interleaving, spacing, and retrieval make you feel like you are struggling. That struggle is often the signal of real learning.

If coaching classes were honest, they would tell students this upfront.

They do not, because it makes the product harder to sell.

A good AI tutor can be honest because it does not need to impress a classroom.

It just needs to make the student better.

References (academic)

• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380. https://doi.org/10.1037/0033-2909.132.3.354
• Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving Students’ Learning With Effective Learning Techniques. Psychological Science in the Public Interest, 14(1), 4–58. https://doi.org/10.1177/1529100612453266
• Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is spacing the “enemy of induction”? Psychological Science, 19(6), 585–592. https://doi.org/10.1111/j.1467-9280.2008.02127.x
• Roediger, H. L., & Karpicke, J. D. (2006). Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science, 17(3), 249–255. https://doi.org/10.1111/j.1467-9280.2006.01693.x
• Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498. https://doi.org/10.1007/s11251-007-9015-8