Artifipedia
GlossaryMapStudyBlogUniversitiesMy Library
Home/Tracks

Learning tracks

Routes with an end

312 concepts is a corpus, not a course. Each track below is a bounded route to one destination: every concept it depends on, in the order they build. Computed from the prerequisite graph, so they stay correct as the encyclopedia grows.

How language models work

From raw text to a working chatbot. The route most people actually want.

24
24 concepts3 fields
  1. 1Training Data
  2. 2Natural Language Processing
  3. 3Supervised Learning
  4. 4Curse of Dimensionality
  5. 5Unsupervised Learning
  6. 6Tokenization
  7. 7Perceptron
  8. 8Regression
  9. 9Dimensionality Reduction
  10. 10Token
  11. 11Neural Network
  12. 12Loss Function
  13. 13Activation Function
  14. 14Embeddings
  15. 15RNN
  16. 16Encoder-Decoder
  17. 17Softmax
  18. 18Layer Normalization
  19. 19Self-Supervised Learning
  20. 20Attention
  21. 21Self-Attention
  22. 22Positional Encoding
  23. 23Transformer
  24. 24Large Language Model

Building with retrieval

Grounding a model in your own documents, and why that is harder than it looks.

28
28 concepts4 fields
  1. 1Training Data
  2. 2Natural Language Processing
  3. 3Supervised Learning
  4. 4Curse of Dimensionality
  5. 5Unsupervised Learning
  6. 6Tokenization
  7. 7Perceptron
  8. 8Regression
  9. 9Dimensionality Reduction
  10. 10Token
  11. 11Chunking
  12. 12Neural Network
  13. 13Loss Function
  14. 14Embeddings
  15. 15Activation Function
  16. 16RNN
  17. 17Encoder-Decoder
  18. 18Softmax
  19. 19Layer Normalization
  20. 20Self-Supervised Learning
  21. 21Vector Search
  22. 22Attention
  23. 23Semantic Search
  24. 24Self-Attention
  25. 25Positional Encoding
  26. 26Transformer
  27. 27Large Language Model
  28. 28Retrieval-Augmented Generation

AI agents

Systems that take actions rather than only produce text.

25
25 concepts4 fields
  1. 1Training Data
  2. 2Natural Language Processing
  3. 3Supervised Learning
  4. 4Curse of Dimensionality
  5. 5Unsupervised Learning
  6. 6Tokenization
  7. 7Perceptron
  8. 8Regression
  9. 9Dimensionality Reduction
  10. 10Token
  11. 11Neural Network
  12. 12Loss Function
  13. 13Activation Function
  14. 14Embeddings
  15. 15RNN
  16. 16Encoder-Decoder
  17. 17Softmax
  18. 18Layer Normalization
  19. 19Self-Supervised Learning
  20. 20Attention
  21. 21Self-Attention
  22. 22Positional Encoding
  23. 23Transformer
  24. 24Large Language Model
  25. 25AI Agent

AI safety and alignment

Why making these systems do what we intend is unsolved.

31
31 concepts5 fields
  1. 1Training Data
  2. 2Artificial Intelligence
  3. 3Natural Language Processing
  4. 4Supervised Learning
  5. 5Curse of Dimensionality
  6. 6Unsupervised Learning
  7. 7Tokenization
  8. 8AI Safety
  9. 9Perceptron
  10. 10Regression
  11. 11Dimensionality Reduction
  12. 12Token
  13. 13Neural Network
  14. 14Loss Function
  15. 15Activation Function
  16. 16Embeddings
  17. 17RNN
  18. 18Transfer Learning
  19. 19Encoder-Decoder
  20. 20Softmax
  21. 21Layer Normalization
  22. 22Self-Supervised Learning
  23. 23Attention
  24. 24Self-Attention
  25. 25Positional Encoding
  26. 26Transformer
  27. 27Large Language Model
  28. 28Fine-tuning
  29. 29Instruction Tuning
  30. 30RLHF
  31. 31AI Alignment

Reasoning models

The longest chain in the corpus: everything behind models that think before answering.

34
34 concepts3 fields
  1. 1Training Data
  2. 2Natural Language Processing
  3. 3Supervised Learning
  4. 4Curse of Dimensionality
  5. 5Unsupervised Learning
  6. 6Tokenization
  7. 7Perceptron
  8. 8Regression
  9. 9Dimensionality Reduction
  10. 10Token
  11. 11Neural Network
  12. 12Loss Function
  13. 13Activation Function
  14. 14Embeddings
  15. 15RNN
  16. 16Transfer Learning
  17. 17Encoder-Decoder
  18. 18Softmax
  19. 19Layer Normalization
  20. 20Self-Supervised Learning
  21. 21Attention
  22. 22Self-Attention
  23. 23Positional Encoding
  24. 24Transformer
  25. 25Large Language Model
  26. 26Prompt Engineering
  27. 27Fine-tuning
  28. 28Instruction Tuning
  29. 29In-Context Learning
  30. 30RLHF
  31. 31Chain-of-Thought
  32. 32Reasoning
  33. 33RLVR
  34. 34Reasoning Model

How these are built. Every concept records what must be understood before it. A track is the full set of prerequisites for one destination, ordered so nothing appears before what it depends on. Nothing here is hand-written curriculum: change the graph and the tracks change.

Progress is read from the concepts you have opened, stored in this browser only. Nothing is gated, and you can jump in anywhere. Want a destination that is not listed? Plan what to learn builds a route to any of the 312 concepts.