Agent skill · research science · aradotso
thereisnospoon-ml-primer
thereisnospoon-ml-primer — an agent skill for coding agents.
Why this skill is useful
Provides domain-specific knowledge and engineering analogies for understanding machine learning concepts that the AI wouldn't reliably generate on its own.
What it needs
Requires matplotlib, numpy installed locally. About 7k tokens when loaded. Last updated 2026-07-09. 70 stars on the source repository.
What this skill does
bash git clone https://github.com/dreddnafious/thereisnospoon.git cd thereisnospoon bash pip install matplotlib numpy bash python3 scripts/01neuronhyperplane.py python3 scripts/02activationfunctions.py python3 scripts/03paperfolding.py python3 scripts/04derivatives.py python3 scripts/05chainrule.py python3 scripts/06attention.py python3 scripts/07ffnvolumetric.py python3 scripts/08residualconnections.py python3 scripts/09dotproducts.py python3 scripts/10losslandscapes.py python3 scripts/11combinationrules.py python3 scripts/12gatingoperations.py bash for f in scripts/.py; do python3 "$f"; done thereisnospoon/ ├── ml-primer.md # The full primer — primary content ├── SYLLABUS.md # Full topic map / table of contents ├── figures/ # SVG/PNG visualizations (auto-generated) │ ├── logo.svg │ ├── 01neuronhyperplane. │ └── ... └── scripts/ # Python figure-generation scripts ├── 01neuronhyperplane.py ├── 02activationfunctions.py └── ... python import numpy as np def neuron(x: np.ndarray, w: np.ndarray, b: float) -> float: """ Single neuron: dot product + bias + nonlinearity. Conceptually: how much does input x align with direction w? """ preactivation = np.dot(w, x) + b # directional agreement return np.maximum(0, preactivation) # ReLU nonlinearity Example: 3-dimensional input x = np.array([0.5, -0.3, 0.8]) w = np.array([1.0, 0.0, 0.5]) # "cares about" dims 0 and 2 b = -0.2 output = neuron(x, w, b) print(f"Neuron output: {output:.4f}") python import numpy as np def denselayer(X: np.ndarray, W: np.ndarray, b: np.ndarray) -> np.ndarray: """ X: (batch, infeatures) W: (infeatures, outfeatures) b: (outfeatures,) Returns: (batch, outfeatures) after ReLU """ return np.maximum(0, X @ W + b) Two-layer MLP: paper folding twice np.random.seed(42) X = np.random.randn(8, 4) # 8 examples, 4 features W1 = np.random.randn(4, 16) 0.1 b1 = np.zeros(16) W2 = np.random.randn(16, 2) 0.1 b2 = np.zeros(2) hidden = denselayer(X, W1, b1) # fold once output = denselayer(hidden, W2, b2) # fold again pri …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/thereisnospoon-ml-primer