Caveman Coding

NO TOOLS. NO MAGIC. HIT THE ROCK YOURSELF.

Why it exists

The classic advice for learning to code was to go and build projects. It worked because building forced you through every small problem yourself. Generative AI quietly broke that: it will happily build the advanced project for you, so people now start with sophisticated work where the AI writes the difficult half, and skip half the learning. You raise a tower without ever lifting a rock.

Caveman Coding was built after finishing a computer science degree and realising that the fundamentals underneath the projects felt scratchy, precisely because modern tooling made it so easy never to practise them.

The method: Fundamental Hybrid Learning

You become the libraries before you use them. Before reaching for a tool, you implement its core idea by hand: the loop, the hash lookup, the traversal, the equation. When you later call the library, you know exactly what you are calling and why.

1. 01 The Entrance Iteration, hashing, prefix sums, two pointers
2. 02 The Deep Passages Recursion, stacks, sorting, graphs, DP
3. 03 The Counting Chamber Primes, counting, Monte Carlo, Bayes
4. 04 The Vector Halls Linear algebra from bare lists
5. 05 The Slope Caves Derivatives, gradient descent, regression
6. 06 The Painted Hall Perceptron, backprop, the letter reading capstone

The rules of the cave

1. No LLMs, no chatbots, no copilots, no autocomplete.
2. No copy and paste. Every line is typed by hand.
3. Official documentation is the only wall you may read.
4. Mathematics is encouraged. Pen and paper are the original IDE.
5. Stuck for fifteen minutes? Open the session’s hint, not a chatbot.

The tracker is a single page that runs straight from the file in a browser, with no server, no dependencies and no build step. Progress is saved locally and can be exported to a plain file and restored on another machine.