The Case For
A New Medium
Or explore at your own pace. ↓
The world education was built for
Knowledge was scarce. One expert. Hundreds of students. The lecture was a solution.
Books could travel. But teachers still couldn't.
Factories needed reliable workers. Schools taught reliability.
How accurately can you reproduce what you received? Memorization as fidelity test.
The signal can now be retrieved by anyone. In seconds. For free. The test no longer tests what it was designed to test.
Why memorization became dominant
Memorization was not a pedagogical philosophy. It was an engineering constraint.
When books cost a fortune and teachers were rare, packing knowledge into human heads was the only available transmission medium.
Examinations were fidelity tests: how accurately can you reproduce what you received? This system was, for its era, a marvel of scale.
India educated 265 million children through it. The United States built a middle class on it. It served its purpose.
The problem is that its purpose has now been automated away.
The AI inflection
November 30, 2022.
OpenAI released ChatGPT.
What took twelve years of schooling, a language model now does in seconds. For free. At global scale.
The economic moat of memorization collapsed overnight.
“But it was never memorization that mattered. Memorization was the proxy — the measurable signal — for the thing educators actually cared about: understanding.”
The collapse of recall-based value
Declining skills
Rising skills
Source: WEF Future of Jobs Report 2025
of core workforce skills will be transformed or obsolete by 2030.
63% of employers cite the skills gap as their #1 barrier to transformation.
The crisis in modern education
“Only 42.6% of Indian graduates were considered employable in 2024.”
Average private schooling cost per child, age 3–17
Engineering graduates expected to find engineering employment in 2024
TeamLeaseStudents who died by suicide in Kota in 2023
Student suicides in India, 2019–2021
Government Parliamentary DataThe sacrifice is real. The return on that sacrifice is collapsing.
This is what happens when a system built for one world tries to prepare people for a different one.
The cognitive science of understanding
The brain doesn't store understanding. It builds it.
Cognitive science has a term for what passive learning produces: the fluency illusion.
When you watch a well-produced explanation, your brain's comprehension circuits fire. It feels like learning. But comprehension is not understanding.
Comprehension is what happens when information passes through your mind without resistance. Understanding is what happens when your mind is permanently restructured by an idea.
Active learning: 90%+ retention. Passive learning: 5–10% retention. This is not a marginal difference. This is the difference between education that works and education that simulates working.
Source: PMC6749092 · Nature Neuroscience
Why current edtech still fails
The first wave of edtech solved the wrong problem.
Coursera. edX. Khan Academy. They democratized content access. This was valuable. But content access was never the binding constraint.
The binding constraint was always the quality of the cognitive process happening inside the learner's mind. Recorded video is a higher-fidelity lecture. It is still passive. The fluency illusion still occurs.
Explorable explanations & constructionism
In 1980, Seymour Papert published Mindstorms. His insight: children learn best not by being taught, but by constructing knowledge through active engagement with objects that embody ideas. He called these “objects to think with.”
“The goal is not to deliver information. The goal is to change the mind.”
— Bret Victor, Explorable Explanations (2011)
The optimal learning medium for a technical concept is not an explanation of the concept. It is a system that embodies the concept, which the learner can probe.
This is the thesis — demonstrated
This is the counter-intuitive result of Bayes' Theorem — high test accuracy alone is not enough.
Why ML is the ideal starting wedge
Not all subjects suffer equally from passive learning.
Machine learning is on the extreme end of conceptual density. To genuinely understand it requires intuition for probability, multivariable calculus, linear algebra, optimization, and generalization.
Each of these is a domain where the gap between knowing the formula and having intuition is enormous — and where that gap matters enormously in practice.
The platform vision
Machine learning is the wedge. Not the ceiling.
Every difficult technical subject — physics, chemistry, economics, neuroscience, probability theory — shares the same structure: a small number of fundamental conceptual insights that, once grasped intuitively, unlock enormous explanatory power.
Passive platforms observe: play, pause, completion rate. Interactive platforms observe: where the learner pauses, which parameter they adjust first, where their mental model breaks, and how it rebuilds. This is a fundamentally richer signal. It compounds with every interaction.
Why Nous exists
One night. 150 students. I had been called in to teach a single ML session in grad school — one night to prepare, no training as a teacher, just someone who understood the subject.
Sitting at my desk that night I realised something uncomfortable. You cannot just throw machine learning concepts at people and expect them to understand. The way it was being taught assumed that if you explained something clearly enough, it would land. It doesn't.
So I built visualisations from scratch. Analogies. Ways of making the invisible visible. I came into that room the next day and tried them on the class.
Something happened I was not expecting. The students weren't just understanding — they were building on the analogies, extending them, using them to figure out things I hadn't even explained yet.
The professors pulled me aside and asked if they could use my slides for their upper classes. One session. 150 students. That was the first night I felt I had accomplished something truly great in my life — not because of the result, but because I could see it happening in real time.
I went home that night and slept better than I had in years. That was the last night I slept peacefully.
Because every night after that, in every gap I could find, I was thinking about how to do it better. More interactivity. More scale. Every subject, not just ML. The feeling of watching that room of 150 students was something I could not let go of.
In India, and in countries like it, education is not optional. It is the thing. Families will starve, skip meals, take loans, work longer hours, sacrifice everything they have — because they believe that if their child gets the best education, everything else will follow.
And here is the question I cannot stop thinking about. Are they actually getting the best?
I launched quietly. Told nobody. On day one, 30 people from 5 countries found it. 57 percent stayed and explored.
That night, someone I had never met sent me a message. His name is Allamaprabhu. He said:
“The first thing I liked was the UI, it's minimal and simple, it's more gen Z which I and the current generation would prefer. The second thing I loved was the context, the narration and content writing is so chill, easy to grasp and creates curiosity to read more and more. Then I loved interacting with those 2D/3D sections and visually coping up.”
I hadn't asked for feedback. He just felt something and had to say it. That is the feeling I am chasing — not for ML students, but for every person who ever gave up on understanding something hard and quietly decided they were not smart enough, when the truth was that the medium had failed them.
Education has been passive for 500 years. I am making it interactive.
The families sacrificing everything deserve a system that actually delivers understanding. Not memorisation. Not the thing AI already does for free. Understanding. That is what I am building.
— N Sai Harshith Varma, Bangalore, 2026
The future medium for understanding
Every major shift in cognitive technology changed not just what we can learn — but how learning optimally happens.
Writing changed memory. The printing press changed literacy. The internet changed information access. Each technology did not improve the old medium. It made a new mode of cognition possible and, eventually, dominant.
We are at such an inflection point now. AI makes possible, for the first time at scale, what cognitive science has always known is optimal but technology could never deliver: interactive, adaptive, simulation-based construction of understanding.
That company is Nous.