::doctorate-grade · self-directed · 12 deep-dives
A way to keep learning more here.
A self-directed PhD-grade AI track that lives on this site forever. Real papers. Real textbooks. Real exercises with observable milestones. Free.
::deep 01
Mathematical Foundations for AI Research
Linear algebra, probability, calculus, and optimization — the load-bearing math under every frontier paper
::deep 02
Foundational Machine Learning
Supervised, unsupervised, and reinforcement learning fundamentals — the canon you need before deep learning
::deep 03
Deep Learning Fundamentals
From the perceptron to ResNets — the canon before transformers
::deep 04
Transformers from Scratch
Attention, the architecture, and the implementation that underlies every frontier model
::deep 05
Training Dynamics and Scaling
Scaling laws, optimizers, mixed precision, gradient accumulation, and the practical art of training large models
::deep 06
RLHF and Alignment
From InstructGPT to Constitutional AI — how raw language models become helpful, harmless, and honest assistants
::deep 07
Mechanistic Interpretability
Reverse-engineering neural networks — circuits, features, and the path to understanding what models actually do
::deep 08
Multimodal Models
Vision-language models, audio, video, and the architectures that bridge modalities
::deep 09
Agents and Tool Use
ReAct, Toolformer, MCP, and the architectures of language models that act in the world
::deep 10
AI Safety (Technical)
From Concrete Problems to mesa-optimization and alignment faking — the technical research agenda for making AI go well
::deep 11
Capability Evaluation
Benchmarks, evals, and the science of measuring what frontier models can actually do
::deep 12
Frontier Research Patterns
Reading papers, replicating results, and developing the meta-skills of an AI researcher