EBO Died. Its Best Idea Didn’t.

A few articles ago, I killed EBO.

I meant it.

EBO is not secretly coming back.

I am not rebuilding the standalone productivity app under another name.

The contradiction that killed it still exists.

A system that wants to understand studying needs evidence.

A standalone tracker does not naturally have enough of it.

So it asks the student to log more.

Or it observes more.

Or it makes shallow conclusions from weak signals.

None of those paths matched what I wanted.

But one idea inside EBO kept bothering me.

Not the tracker.

Not the social layer.

Not the dashboards.

The idea that a system could help a student understand their preparation rather than merely record activity.

That idea survived.

It just needed a better place to live.

EBO was standing outside the studying

Imagine I study from a book for ninety minutes.

EBO can know:

I started a timer.

I ended a timer.

Maybe I selected Quant.

Maybe I selected Algebra.

That is still very little.

What happened inside those ninety minutes?

How many questions?

Which skills?

What difficulty?

What accuracy?

What mistakes?

Did I improve?

Was I slow?

Did I guess?

Did I repair an earlier weakness?

The tracker does not know.

So it asks me.

And now the productivity app creates productivity paperwork.

That is exactly the problem I stopped wanting to solve.

Verbal was different without trying to be

Inside Verbal, the learning itself creates evidence.

If I review a word, the app knows which word.

If I fail recall, it knows.

If I confuse two words, it knows.

If I get something right after several lapses, it knows the history.

If I answer a verbal question, it knows the topic, timing and result.

No post-session form required.

The evidence appears because the product needed it to perform the learning interaction.

That is a much healthier relationship with data.

Quant made the difference impossible to ignore

Quant makes this even clearer.

A solve can naturally produce:

Question identity.

Topic.

Subtopic.

Difficulty.

Format.

Answer.

First answer.

Final answer.

Correctness.

Timing.

Confidence.

Mistake classification.

Prior exposure.

Review state.

Later repair.

Again:

No academic tax return after the session.

The student is already doing the thing the product needs to understand.

That means the intelligence can be built from activity rather than from reporting activity.

That is the idea EBO was reaching for

EBO wanted to answer questions like:

What are you weak at?

What are you improving at?

How consistent are you?

Where are you wasting effort?

What should probably come next?

Those are reasonable questions.

The mistake was assuming a separate tracker should be the place that answered them.

Now the subject products can produce the evidence directly.

That changes the problem.

Intelligence can sit above the work instead of beside it

The new structure is cleaner.

Verbal handles Verbal.

Quant handles Quant.

A separate Intelligence layer can look across the evidence both produce.

The learner does not need another place to maintain progress manually.

The intelligence system can remain mostly invisible until there is something useful to say.

That is much closer to what I wanted EBO to feel like.

Helpful without becoming another job.

The privacy problem changes too

EBO became uncomfortable partly because the smarter it wanted to become, the more context it seemed to need from the student's wider life.

The new system can stay much narrower.

It does not need to know everything I did today.

It needs to understand what happened inside the learning products.

That is still personal data.

It still needs careful boundaries.

But there is a big difference between:

We recorded the learning interaction you just performed

and:

We need to monitor more of your life so we can guess whether you were productive.

The first is much easier to justify.

Progress can become academic instead of performative

A study tracker naturally gravitates toward visible activity.

Hours.

Sessions.

Streaks.

Tasks.

Those can be useful.

They are also easy to optimise without learning.

The subject apps have access to better signals.

A mistake repaired later.

A word remembered after a longer interval.

A timed weakness improving.

A concept transferring to a new question.

A mock-selection problem disappearing.

Those are closer to the thing I care about.

Consistency can remain one signal among many.

Not the whole definition of progress.

This does not make the system omniscient

If I study outside the apps, the system may not know.

That is fine.

I do not need to solve that perfectly.

A product does not need total knowledge to be useful.

It needs to be honest about the knowledge it has.

If I solve a worksheet on paper, Intelligence should not pretend it observed that work.

If I use Quant for thirty questions, it can reason about those thirty questions.

That narrower honesty is more valuable than fake completeness.

I am not rebuilding a universal life tracker

That boundary matters.

The new intelligence system is about preparation evidence.

Skills.

Performance.

Retention.

Timing.

Mocks.

Review.

Learning behaviour inside the products.

It is not trying to infer whether every hour of my day was productive.

I do not want that.

The narrower scope is a feature.

It gives the system better evidence and fewer excuses to become invasive.

EBO still changed how I design everything

Even dead projects leave habits behind.

EBO made me suspicious of:

Manual logging.

Vanity metrics.

Punitive streaks.

Social pressure.

Confident conclusions from incomplete data.

Progress systems that become another obligation.

Those suspicions now shape Verbal, Quant and Intelligence.

So EBO did not survive as a product.

It survived as constraints.

That may be more useful.

The best idea was not the dashboard

Looking back, the most interesting idea in EBO was never:

Show me my productivity.

It was:

Help me understand what my behaviour suggests and what I should do next.

That is a much stronger idea when the behaviour comes from real learning evidence.

Now a recommendation can be based on actual misses.

Actual retention.

Actual speed.

Actual mock behaviour.

Actual review debt.

Not just hours and self-reported categories.

That changes the question from:

Did I study enough?

to:

What part of my preparation deserves attention next?

I care much more about the second one.

A dead product can still be useful

If I had not pushed EBO far enough to find the contradiction, I might have built the same tracking layer again inside the IPMAT project.

Instead, I recognised the smell.

Too much manual context.

Too much pressure to interpret weak signals.

Too much temptation to turn studying into managing the studying system.

So when Verbal and Quant started producing better evidence naturally, I could see the difference immediately.

EBO failed.

The failure made the next architecture better.

There is something satisfying about this ending

The original EBO vision was:

Make studying easier to begin, sustain and return to.

The standalone app began undermining that purpose.

So I killed it.

Now some of the useful intelligence ideas can exist in a system where the student does not have to maintain another productivity layer at all.

That feels like a much better resolution than pretending the project never mattered.

The product died.

The constraint stayed.

And the best idea found a place where it finally makes sense.