Cheat sheets

Topics

1. APA2A ProtocolDiscover an agent, send it work, follow the task, read its errors. Pinned to A2A 1.0.1.40 rows · 2 diagrams · 13 blocks
2. AHAgent Harness EngineeringThe code around the model. Loop contract, hooks, budgets, tools, gates, sandbox, traces and workbench files.33 rows · 2 diagrams · 13 blocks
3. ASAI Safety and Red TeamingName the attack, detect it, measure refusals, gate the output and audit alignment. Phase 18 and 19 lessons.30 rows · 2 diagrams · 12 blocks
4. AAAutonomous Agents and PermissionsRun an agent unattended with limits. Permission modes, budgets, kill switches, durable execution, approvals, rollback.24 rows · 2 diagrams · 12 blocks
5. CMClassical Machine LearningFit, validate, tune and fix regression, classification and clustering models with scikit-learn.30 rows · 2 diagrams · 12 blocks
6. CVComputer VisionPreprocess, convolve, classify, fine-tune, detect, segment and patchify images. PyTorch and NumPy from the lessons.37 rows · 2 diagrams · 12 blocks
7. TNTraining Neural NetworksPick the optimizer, init, normalization, regularization and schedule. Then debug the run when the loss misbehaves.40 rows · 2 diagrams · 12 blocks
8. DGDiffusion and Generative ModelsTrain and sample a VAE, a GAN, DDPM, latent diffusion and flow matching, then measure them. Toy Python from the lessons.32 rows · 2 diagrams · 11 blocks
9. DSDocker Sandboxes for AgentsRun an agent in a microVM, control its network and keys, get its commits back. sbx 0.47.0, docker-agent 1.149.0.35 rows · 2 diagrams · 12 blocks
10. EVEmbeddings and Vector SearchEncode text, pick a metric, size the vectors, choose an index and fix bad recall. Python, NumPy and FAISS.31 rows · 2 diagrams · 12 blocks
11. IOInference Optimization: KV Cache, Quantization, BatchingSize the KV cache, batch continuously, pick a quantization format, and measure what users feel.29 rows · 2 diagrams · 12 blocks
12. ELEvaluating LLMsSpecify tasks, score them, judge with a rubric, calibrate, rank with intervals and test long context.35 rows · 2 diagrams · 12 blocks
13. FTFine-Tuning with LoRAAdd LoRA adapters, load a 4-bit base for QLoRA, mask SFT labels, train, merge and debug. PyTorch, PEFT and bitsandbytes.28 rows · 2 diagrams · 12 blocks
14. MMMath for Machine LearningGradients, probability, information, optimizers, decompositions and float safety. NumPy, SciPy and PyTorch.32 rows · 2 diagrams · 12 blocks
15. MCModel Context Protocol (MCP)Build, call and serve stateless MCP 2026-07-28 servers over stdio and Streamable HTTP.32 rows · 2 diagrams · 12 blocks
16. MAMulti-Agent SystemsAgents, handoffs, shared state and orchestrators. Pick a pattern, vote, checkpoint and fix the known failures.31 rows · 2 diagrams · 12 blocks
17. PEPrompt EngineeringWrite the prompt, add examples and reasoning, vote, and budget the context window. Python from the lessons.32 rows · 2 diagrams · 11 blocks
18. PTPyTorch EssentialsTensors, autograd, nn.Module, the training loop, DataLoader and devices, from the lessons.44 rows · 2 diagrams · 12 blocks
19. RARetrieval-Augmented Generation (RAG)Chunk, index, retrieve, fuse, rerank, generate with citations and grade each stage. Python from the lessons.32 rows · 2 diagrams · 12 blocks
20. RLReinforcement LearningMDPs, value methods, policy gradients, PPO, RLHF and GRPO. Formulas, update lines and diagnostics from Phase 9.32 rows · 2 diagrams · 12 blocks
21. REReverse Engineering with reaRead a compiled program and prove what you found. Pinned to rea-agents 6.5.0 with Ghidra 12.1.4.48 rows · 2 diagrams · 13 blocks
22. LSLLM Serving with SGLangStart, query, size, measure and fix an SGLang server. Pinned to SGLang 0.5.22.40 rows · 2 diagrams · 13 blocks
23. TBTokenization and BPETrain, encode and decode a byte-level BPE tokenizer, then slice the id stream into training windows.24 rows · 2 diagrams · 11 blocks
24. TCTool Calling and Structured OutputsDeclare tools, read calls, return results, run calls in parallel, stream arguments, and force JSON that validates.29 rows · 2 diagrams · 13 blocks
25. TATransformers and AttentionCompute attention, mask it, split heads, add positions and stack the block. PyTorch and NumPy from the lessons.32 rows · 2 diagrams · 11 blocks

By phase

• Phase 0: Setup & ToolingPyTorch Essentials
• Phase 1: Math FoundationsEmbeddings and Vector Search, Math for Machine Learning
• Phase 2: ML FundamentalsClassical Machine Learning, Embeddings and Vector Search
• Phase 3: Deep Learning CoreTraining Neural Networks, PyTorch Essentials
• Phase 4: Computer VisionComputer Vision, Embeddings and Vector Search
• Phase 5: NLP, Foundations to AdvancedEmbeddings and Vector Search, Evaluating LLMs, Retrieval-Augmented Generation (RAG), Tokenization and BPE, Transformers and Attention
• Phase 7: Transformers Deep DiveInference Optimization: KV Cache, Quantization, Batching, Transformers and Attention
• Phase 8: Generative AIDiffusion and Generative Models
• Phase 9: Reinforcement LearningReinforcement Learning
• Phase 10: LLMs from ScratchInference Optimization: KV Cache, Quantization, Batching, Evaluating LLMs, Fine-Tuning with LoRA, Tokenization and BPE, Transformers and Attention
• Phase 11: LLM EngineeringEmbeddings and Vector Search, Evaluating LLMs, Fine-Tuning with LoRA, Prompt Engineering, Retrieval-Augmented Generation (RAG), Tool Calling and Structured Outputs
• Phase 13: Tools & ProtocolsModel Context Protocol (MCP), Reverse Engineering with rea, Tool Calling and Structured Outputs
• Phase 14: Agent EngineeringAgent Harness Engineering
• Phase 15: Autonomous SystemsAgent Harness Engineering, Autonomous Agents and Permissions
• Phase 16: Multi-Agent & SwarmsMulti-Agent Systems
• Phase 17: Infrastructure & ProductionInference Optimization: KV Cache, Quantization, Batching
• Phase 18: Ethics, Safety & AlignmentAI Safety and Red Teaming
• Phase 19: Capstone ProjectsAgent Harness Engineering, AI Safety and Red Teaming, Embeddings and Vector Search, Evaluating LLMs, Fine-Tuning with LoRA, Retrieval-Augmented Generation (RAG), Tokenization and BPE, Tool Calling and Structured Outputs, Transformers and Attention

How to use them

• Each colour means one kind of block. Blue blocks are tasks and commands. Red blocks are fixes. The key is at the top of each poster.
• Click a command to copy it. Replace each value in angle brackets, such as <model>.
• Each block names the lesson or manual section that its facts come from. Open it when you need the reason.
• Print a poster on one or two landscape pages. It fits A4 and Letter paper.