https://ai.runbook.academy/https://ai.runbook.academy/about/https://ai.runbook.academy/capstone/https://ai.runbook.academy/capstone/capstone-agentic-workstation/https://ai.runbook.academy/capstone/capstone-cluster-or-tiered-deployment/https://ai.runbook.academy/capstone/capstone-hardware-and-model-plan/https://ai.runbook.academy/capstone/capstone-improve-a-small-model/https://ai.runbook.academy/capstone/capstone-inference-service/https://ai.runbook.academy/capstone/capstone-overview/https://ai.runbook.academy/capstone/capstone-report/https://ai.runbook.academy/curriculum/https://ai.runbook.academy/how-to-use-this-course/https://ai.runbook.academy/part-01-machine-learning-foundations/https://ai.runbook.academy/part-01-machine-learning-foundations/generalisation-and-overfitting/https://ai.runbook.academy/part-01-machine-learning-foundations/lab-python-environment-and-first-model/https://ai.runbook.academy/part-01-machine-learning-foundations/neural-networks-and-backpropagation/https://ai.runbook.academy/part-01-machine-learning-foundations/tensors-gpus-and-precision/https://ai.runbook.academy/part-01-machine-learning-foundations/what-learning-means/https://ai.runbook.academy/part-02-what-a-language-model-is/https://ai.runbook.academy/part-02-what-a-language-model-is/attention-and-the-transformer/https://ai.runbook.academy/part-02-what-a-language-model-is/embeddings-and-vector-space/https://ai.runbook.academy/part-02-what-a-language-model-is/lab-look-inside-a-model/https://ai.runbook.academy/part-02-what-a-language-model-is/next-token-prediction/https://ai.runbook.academy/part-02-what-a-language-model-is/parameters-layers-and-model-size/https://ai.runbook.academy/part-02-what-a-language-model-is/tokens-and-tokenisers/https://ai.runbook.academy/part-03-how-models-are-made/https://ai.runbook.academy/part-03-how-models-are-made/inference-prefill-decode-and-bandwidth/https://ai.runbook.academy/part-03-how-models-are-made/making-small-models-good/https://ai.runbook.academy/part-03-how-models-are-made/open-weights-licences-and-the-law/https://ai.runbook.academy/part-03-how-models-are-made/post-training/https://ai.runbook.academy/part-03-how-models-are-made/pretraining/https://ai.runbook.academy/part-03-how-models-are-made/reality-check-small-local-versus-frontier/https://ai.runbook.academy/part-04-open-weight-landscape/https://ai.runbook.academy/part-04-open-weight-landscape/choosing-a-model-for-a-memory-budget/https://ai.runbook.academy/part-04-open-weight-landscape/dense-moe-and-hybrid-architectures/https://ai.runbook.academy/part-04-open-weight-landscape/lab-build-your-model-shortlist/https://ai.runbook.academy/part-04-open-weight-landscape/model-cards-and-benchmarks/https://ai.runbook.academy/part-04-open-weight-landscape/model-families-and-who-makes-them/https://ai.runbook.academy/part-04-open-weight-landscape/reading-a-model-name/https://ai.runbook.academy/part-05-your-hardware/https://ai.runbook.academy/part-05-your-hardware/amd-ryzen-ai-max-strix-halo/https://ai.runbook.academy/part-05-your-hardware/apple-silicon-unified-memory/https://ai.runbook.academy/part-05-your-hardware/lab-measure-bandwidth-and-compute/https://ai.runbook.academy/part-05-your-hardware/lab-prepare-your-machine/https://ai.runbook.academy/part-05-your-hardware/nvidia-desktops-and-laptops/https://ai.runbook.academy/part-05-your-hardware/nvidia-dgx-spark/https://ai.runbook.academy/part-05-your-hardware/why-memory-not-flops/https://ai.runbook.academy/part-06-llama-cpp/https://ai.runbook.academy/part-06-llama-cpp/challenge-two-tokens-per-second/https://ai.runbook.academy/part-06-llama-cpp/gguf-and-quantisation-types/https://ai.runbook.academy/part-06-llama-cpp/installing-and-building-llama-cpp/https://ai.runbook.academy/part-06-llama-cpp/lab-benchmark-the-reference-models/https://ai.runbook.academy/part-06-llama-cpp/llama-cli-and-llama-server/https://ai.runbook.academy/part-06-llama-cpp/llama-cpp-the-engine-that-runs-everywhere/https://ai.runbook.academy/part-06-llama-cpp/sampling-and-determinism/https://ai.runbook.academy/part-07-ollama-and-lm-studio/https://ai.runbook.academy/part-07-ollama-and-lm-studio/front-ends-open-webui-and-friends/https://ai.runbook.academy/part-07-ollama-and-lm-studio/lab-private-chat-service-on-your-lan/https://ai.runbook.academy/part-07-ollama-and-lm-studio/lm-studio-gui-mlx-and-headless/https://ai.runbook.academy/part-07-ollama-and-lm-studio/managing-a-model-library/https://ai.runbook.academy/part-07-ollama-and-lm-studio/ollama-models-as-a-service/https://ai.runbook.academy/part-07-ollama-and-lm-studio/reality-check-the-default-context-is-enough/https://ai.runbook.academy/part-08-platform-native-engines/https://ai.runbook.academy/part-08-platform-native-engines/lab-same-model-every-engine/https://ai.runbook.academy/part-08-platform-native-engines/mlx-and-mlx-lm/https://ai.runbook.academy/part-08-platform-native-engines/rocm-builds-lemonade-and-the-npu/https://ai.runbook.academy/part-08-platform-native-engines/specialist-engines/https://ai.runbook.academy/part-08-platform-native-engines/tensorrt-llm-nim-and-the-spark-playbooks/https://ai.runbook.academy/part-09-vllm-and-sglang/https://ai.runbook.academy/part-09-vllm-and-sglang/batching-paged-attention-and-throughput/https://ai.runbook.academy/part-09-vllm-and-sglang/installing-vllm-on-each-platform/https://ai.runbook.academy/part-09-vllm-and-sglang/lab-serve-twenty-concurrent-users/https://ai.runbook.academy/part-09-vllm-and-sglang/project-your-local-model-gateway/https://ai.runbook.academy/part-09-vllm-and-sglang/server-side-tool-calling-and-structured-output/https://ai.runbook.academy/part-09-vllm-and-sglang/serving-with-vllm/https://ai.runbook.academy/part-09-vllm-and-sglang/sglang-when-to-choose-it/https://ai.runbook.academy/part-09-vllm-and-sglang/speculative-decoding/https://ai.runbook.academy/part-10-models-at-work/https://ai.runbook.academy/part-10-models-at-work/lab-benchmark-models-on-your-own-tasks/https://ai.runbook.academy/part-10-models-at-work/local-coding-assistants-in-your-editor/https://ai.runbook.academy/part-10-models-at-work/multimodal-vision-speech-and-documents/https://ai.runbook.academy/part-10-models-at-work/privacy-security-and-serving-beyond-localhost/https://ai.runbook.academy/part-10-models-at-work/project-private-document-qa/https://ai.runbook.academy/part-10-models-at-work/prompting-chat-templates-and-thinking-modes/https://ai.runbook.academy/part-10-models-at-work/retrieval-augmented-generation/https://ai.runbook.academy/part-10-models-at-work/structured-output-and-json-mode/https://ai.runbook.academy/part-11-training-toolchain/https://ai.runbook.academy/part-11-training-toolchain/datasets-templates-tokenisation-and-packing/https://ai.runbook.academy/part-11-training-toolchain/experiment-tracking-and-reproducibility/https://ai.runbook.academy/part-11-training-toolchain/lab-your-first-training-run/https://ai.runbook.academy/part-11-training-toolchain/memory-arithmetic-for-training/https://ai.runbook.academy/part-11-training-toolchain/pytorch-transformers-and-the-hugging-face-ecosystem/https://ai.runbook.academy/part-11-training-toolchain/training-environment-on-each-platform/https://ai.runbook.academy/part-12-pretraining-from-scratch/https://ai.runbook.academy/part-12-pretraining-from-scratch/anatomy-of-a-training-loop/https://ai.runbook.academy/part-12-pretraining-from-scratch/data-for-pretraining-and-tokeniser-training/https://ai.runbook.academy/part-12-pretraining-from-scratch/lab-train-a-small-model-in-an-afternoon/https://ai.runbook.academy/part-12-pretraining-from-scratch/project-a-domain-micro-model/https://ai.runbook.academy/part-12-pretraining-from-scratch/scaling-laws-and-home-compute-budgets/https://ai.runbook.academy/part-12-pretraining-from-scratch/what-pretraining-teaches/https://ai.runbook.academy/part-13-supervised-fine-tuning/https://ai.runbook.academy/part-13-supervised-fine-tuning/building-an-sft-dataset/https://ai.runbook.academy/part-13-supervised-fine-tuning/challenge-the-fine-tune-that-got-worse/https://ai.runbook.academy/part-13-supervised-fine-tuning/fine-tuning-with-trl-and-peft/https://ai.runbook.academy/part-13-supervised-fine-tuning/lab-fine-tune-a-small-model-to-your-format/https://ai.runbook.academy/part-13-supervised-fine-tuning/lora-and-qlora-explained/https://ai.runbook.academy/part-13-supervised-fine-tuning/merging-exporting-and-quantising-a-fine-tune/https://ai.runbook.academy/part-13-supervised-fine-tuning/project-a-specialist-assistant/https://ai.runbook.academy/part-13-supervised-fine-tuning/unsloth-axolotl-llama-factory-and-mlx-lm/https://ai.runbook.academy/part-13-supervised-fine-tuning/what-fine-tuning-changes/https://ai.runbook.academy/part-14-preference-and-rl/https://ai.runbook.academy/part-14-preference-and-rl/dpo-and-its-family/https://ai.runbook.academy/part-14-preference-and-rl/from-imitation-to-preferences/https://ai.runbook.academy/part-14-preference-and-rl/grpo-explained/https://ai.runbook.academy/part-14-preference-and-rl/lab-dpo-your-model-to-prefer-your-style/https://ai.runbook.academy/part-14-preference-and-rl/lab-grpo-on-a-verifiable-task/https://ai.runbook.academy/part-14-preference-and-rl/reality-check-rl-makes-small-models-reason/https://ai.runbook.academy/part-14-preference-and-rl/reward-functions-and-verifiers/https://ai.runbook.academy/part-14-preference-and-rl/the-rl-toolchain/https://ai.runbook.academy/part-15-distillation/https://ai.runbook.academy/part-15-distillation/challenge-the-student-that-learned-the-teachers-mistakes/https://ai.runbook.academy/part-15-distillation/lab-logit-distillation-with-trl/https://ai.runbook.academy/part-15-distillation/lab-sequence-level-distillation/https://ai.runbook.academy/part-15-distillation/logit-sequence-and-on-policy-distillation/https://ai.runbook.academy/part-15-distillation/project-the-distillation-pipeline/https://ai.runbook.academy/part-15-distillation/pruning-and-prune-and-distil/https://ai.runbook.academy/part-15-distillation/reasoning-distillation-from-teacher-traces/https://ai.runbook.academy/part-15-distillation/synthetic-data-with-a-local-teacher/https://ai.runbook.academy/part-15-distillation/why-distillation-works/https://ai.runbook.academy/part-16-quantisation-and-evaluation/https://ai.runbook.academy/part-16-quantisation-and-evaluation/challenge-the-benchmark-that-lied/https://ai.runbook.academy/part-16-quantisation-and-evaluation/evaluation-harnesses/https://ai.runbook.academy/part-16-quantisation-and-evaluation/lab-quantise-five-ways-and-measure/https://ai.runbook.academy/part-16-quantisation-and-evaluation/lab-run-a-standard-benchmark-suite/https://ai.runbook.academy/part-16-quantisation-and-evaluation/llm-as-judge-contamination-and-reporting/https://ai.runbook.academy/part-16-quantisation-and-evaluation/measuring-quantisation-damage/https://ai.runbook.academy/part-16-quantisation-and-evaluation/post-training-quantisation-in-depth/https://ai.runbook.academy/part-16-quantisation-and-evaluation/qat-fp8-nvfp4-and-mxfp4/https://ai.runbook.academy/part-17-faster-inference/https://ai.runbook.academy/part-17-faster-inference/lab-train-and-deploy-a-draft-model/https://ai.runbook.academy/part-17-faster-inference/prefix-caching-and-kv-reuse/https://ai.runbook.academy/part-17-faster-inference/speculative-decoding-acceptance-and-payoff/https://ai.runbook.academy/part-17-faster-inference/training-a-draft-model/https://ai.runbook.academy/part-18-parallelism-and-networks/https://ai.runbook.academy/part-18-parallelism-and-networks/lab-build-and-measure-your-cluster-network/https://ai.runbook.academy/part-18-parallelism-and-networks/networking-for-home-clusters/https://ai.runbook.academy/part-18-parallelism-and-networks/tensor-pipeline-expert-data-and-sequence-parallelism/https://ai.runbook.academy/part-18-parallelism-and-networks/the-memory-wall-and-the-bandwidth-wall/https://ai.runbook.academy/part-18-parallelism-and-networks/the-reference-cluster-and-the-single-machine-path/https://ai.runbook.academy/part-19-llama-cpp-rpc/https://ai.runbook.academy/part-19-llama-cpp-rpc/challenge-the-cluster-slower-than-one-machine/https://ai.runbook.academy/part-19-llama-cpp-rpc/lab-a-mixed-platform-cluster/https://ai.runbook.academy/part-19-llama-cpp-rpc/lab-a-model-bigger-than-any-one-machine/https://ai.runbook.academy/part-19-llama-cpp-rpc/llama-cpp-rpc-layers-across-machines/https://ai.runbook.academy/part-19-llama-cpp-rpc/rdma-transport-and-tuning/https://ai.runbook.academy/part-20-nvidia-clusters/https://ai.runbook.academy/part-20-nvidia-clusters/connecting-two-dgx-sparks/https://ai.runbook.academy/part-20-nvidia-clusters/lab-serve-a-400b-class-model-on-two-sparks/https://ai.runbook.academy/part-20-nvidia-clusters/multi-gpu-desktops/https://ai.runbook.academy/part-20-nvidia-clusters/tensorrt-llm-and-dynamo-on-spark-pairs/https://ai.runbook.academy/part-20-nvidia-clusters/vllm-multi-node-with-ray/https://ai.runbook.academy/part-21-apple-clusters/https://ai.runbook.academy/part-21-apple-clusters/exo-automatic-partitioning-across-macs/https://ai.runbook.academy/part-21-apple-clusters/lab-two-mac-cluster-over-thunderbolt-5/https://ai.runbook.academy/part-21-apple-clusters/mlx-distributed-backends/https://ai.runbook.academy/part-21-apple-clusters/reality-check-macs-replace-a-gpu-server/https://ai.runbook.academy/part-22-disaggregated-serving/https://ai.runbook.academy/part-22-disaggregated-serving/lab-kv-cache-offload-and-sharing/https://ai.runbook.academy/part-22-disaggregated-serving/lab-two-machine-prefill-decode/https://ai.runbook.academy/part-22-disaggregated-serving/project-a-tiered-inference-architecture/https://ai.runbook.academy/part-22-disaggregated-serving/sglang-pd-and-nvidia-dynamo/https://ai.runbook.academy/part-22-disaggregated-serving/the-kv-cache-as-a-transferable-object/https://ai.runbook.academy/part-22-disaggregated-serving/vllm-disaggregated-prefill/https://ai.runbook.academy/part-22-disaggregated-serving/why-prefill-and-decode-want-different-hardware/https://ai.runbook.academy/part-23-operating-a-local-ai-service/https://ai.runbook.academy/part-23-operating-a-local-ai-service/backup-upgrades-and-reproducibility/https://ai.runbook.academy/part-23-operating-a-local-ai-service/capacity-planning-and-cost-per-million-tokens/https://ai.runbook.academy/part-23-operating-a-local-ai-service/challenge-the-three-am-oom/https://ai.runbook.academy/part-23-operating-a-local-ai-service/lab-dashboards-for-your-cluster/https://ai.runbook.academy/part-23-operating-a-local-ai-service/observability-for-llm-serving/https://ai.runbook.academy/part-23-operating-a-local-ai-service/routing-and-model-management/https://ai.runbook.academy/part-23-operating-a-local-ai-service/security-for-exposed-endpoints-and-model-supply-chain/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/context-engineering-memory-and-compaction/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/function-calling-end-to-end/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/lab-a-minimal-agent-from-scratch/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/lab-write-and-connect-an-mcp-server/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/model-context-protocol/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/reasoning-models-in-agent-loops/https://ai.runbook.academy/part-24-tools-mcp-and-the-agent-loop/what-an-agent-is/https://ai.runbook.academy/part-25-coding-agents/https://ai.runbook.academy/part-25-coding-agents/aider-with-local-models/https://ai.runbook.academy/part-25-coding-agents/challenge-the-agent-that-escaped-the-sandbox/https://ai.runbook.academy/part-25-coding-agents/claude-code-with-a-local-endpoint/https://ai.runbook.academy/part-25-coding-agents/codex-cli-and-opencode/https://ai.runbook.academy/part-25-coding-agents/editor-agents-cline-kilo-code-continue-and-others/https://ai.runbook.academy/part-25-coding-agents/goose-and-openhands/https://ai.runbook.academy/part-25-coding-agents/lab-one-task-six-agents/https://ai.runbook.academy/part-25-coding-agents/lab-sandbox-your-agent/https://ai.runbook.academy/part-25-coding-agents/project-a-local-agentic-coding-workstation/https://ai.runbook.academy/part-25-coding-agents/the-landscape-of-coding-agents/https://ai.runbook.academy/part-25-coding-agents/which-local-models-can-drive-an-agent/https://ai.runbook.academy/part-26-building-agent-systems/https://ai.runbook.academy/part-26-building-agent-systems/agent-frameworks-compared/https://ai.runbook.academy/part-26-building-agent-systems/agentic-rag-and-research-agents/https://ai.runbook.academy/part-26-building-agent-systems/evaluating-agents/https://ai.runbook.academy/part-26-building-agent-systems/multi-agent-patterns/https://ai.runbook.academy/part-26-building-agent-systems/project-a-multi-agent-system-on-your-cluster/https://ai.runbook.academy/part-26-building-agent-systems/reality-check-agents-are-just-loops/https://ai.runbook.academy/part-26-building-agent-systems/safety-prompt-injection-permissions-and-human-in-the-loop/https://ai.runbook.academy/part-27-improving-models-for-agents/https://ai.runbook.academy/part-27-improving-models-for-agents/collecting-trajectories-from-your-agents/https://ai.runbook.academy/part-27-improving-models-for-agents/distilling-a-big-agent-model-into-a-small-one/https://ai.runbook.academy/part-27-improving-models-for-agents/fine-tuning-for-tool-use-and-your-codebase/https://ai.runbook.academy/part-27-improving-models-for-agents/lab-fine-tune-on-your-own-trajectories/https://ai.runbook.academy/part-27-improving-models-for-agents/rl-on-agent-tasks/https://ai.runbook.academy/progress/https://ai.runbook.academy/reference/ai-problem-solving/https://ai.runbook.academy/reference/commands/https://ai.runbook.academy/reference/disclaimer/https://ai.runbook.academy/reference/downloads/https://ai.runbook.academy/reference/glossary/https://ai.runbook.academy/reference/hardware/https://ai.runbook.academy/reference/lab-execution/https://ai.runbook.academy/reference/models/https://ai.runbook.academy/reference/sources/https://ai.runbook.academy/what-you-need/