I am building Derive, a small attempt to think about learning differently.
Not as a course platform. Not as an encyclopedia. Not as a pile of articles sitting next to each other, pretending knowledge is flat.
The idea is simpler: when you want to understand something complex, you often do not fail because the topic itself is impossible. You fail because of an invisible prerequisite underneath it.
You want to understand a mortgage calculator, but maybe the calculator is not the real problem. Maybe the missing piece is percentages, interest, present value, or annuities.
You want to understand ChatGPT, but underneath that are language, probability, functions, vectors, optimization, neural networks, and transformers.
You want to understand nuclear reactors, but before that you need atoms, neutrons, chain reactions, criticality, heat, turbines, and generators.
Derive is meant to make those layers visible.
The problem with normal learning
Most learning products start at the target:
“Here is a course about X.”
That sounds reasonable, but it is often too late. If the foundations are missing, the course quickly feels harder than it really is. Not because you are too slow, but because the system does not show which building block is missing.
This is an old problem. School, university, videos, encyclopedias, online courses: there is knowledge everywhere. But it is rare to see clearly which concepts a concept depends on.
Knowledge is not a stack of pages. Knowledge is a graph.
The actual product idea
Derive starts with one question:
What do you want to understand?
Then it should not just produce an explanation. It should show the shortest path from what you already know to what you want to understand.
So instead of a generic article, you get something more like:
Number -> Percentage -> Interest -> Present Value -> Annuity -> Monthly PaymentOr:
Language -> Probability -> Function -> Vector -> Optimization -> Neural Network -> TransformerThe interesting part is not only the individual path. It gets interesting when the foundations repeat.
Once you really understand Rate, Feedback Loop, Probability, Function, Energy, Incentive, Model, Network, and a few other primitives, you can suddenly absorb many new topics much faster.
That is the real leverage.
Not “learn 10,000 topics.”
Instead: build a repertoire of reusable first principles that lets you cover 70 percent of many new topics before you properly start.
Why this is interesting to me
I like products that change a mental model.
Derive should not only display content. It should create a specific feeling:
Ah, I do not understand this topic yet. I am just missing this one concept underneath it.
That is a small difference, but an important one. It removes some of the fog from learning. Instead of “everything is too hard”, it becomes “I need to review these three foundations.”
That is much more manageable.
What the first prototype can do
The current version is deliberately small and static. It has predefined target topics, concept pages, learning paths, and a topic library. The data is hand-curated, not magical.
But the product shape is already visible:
- Pick a target topic
- See the prerequisite chain
- Check off missing foundations
- Understand what the topic unlocks
- Learn individual concepts through essence, derivation, examples, and unlocks
I have also started adding many more topics. Not as isolated articles, but along recurring principle stacks. Many topics share the same foundations. That is exactly what Derive should eventually exploit.
What it is not
Derive is not meant to become a normal search engine for knowledge.
It is also not an “AI explains everything” chatbot that sounds nice but does not show what the explanation is built on.
And it is not a course library where every topic starts from zero again.
The vision is closer to a personal knowledge map layer: a system that knows which concepts you already understand, which ones are missing, and which new topics become easier because of that.
What is still missing
The first big step is product clarity. The interface has to show immediately why Derive is different. Less wall of text, more visual understanding. More graph, more derivation, more “I can see what the problem is now.”
After that, it gets interesting:
- Persistent progress
- A real repertoire of known principles
- Automatic decomposition of new topics
- Better visual concept pages
- Exercises that make you derive instead of only recall
- Sources, reviews, and quality control
I want Derive to eventually feel like a learning system for people who actually want to understand how things connect.
Not consume faster.
See through things faster.
No link yet. No launch theater yet. I first want the thing to really click as a product.