• Activation functions in LLMs: SwiGLU vs GELU vs ReLUWhy do modern LLMs use SwiGLU and GELU instead of ReLU? Discover the mathematics, gating mechanisms, and the impact on compute and memory.
• Becoming an AI agent engineer in 2026: from prompt to productionDemand for AI agent engineers grew 280% in a year. Learn what the job involves, which skills you need and how to get agents into production
• AI and LLM Terms from A to Z: The Complete Glossary forDiscover the most important AI and LLM terms in this complete glossary for beginners. From RAG and tokens to fine-tuning, explained simply and clearly.
• Alignment explained: how AI models are tunedHow are raw LLMs made safe and helpful? Read all about alignment, RLHF, DPO, constitutional AI, and the balance between helpfulness and safety.
• The Attention Mechanism in AI Explained Simply - llmnet.nlLearn how the attention mechanism in LLMs works. A clear explanation of context, query/key/value, self-attention, and quadratic scaling without heavy math.
• Catastrophic forgetting when fine-tuning LLMsUnderstand catastrophic forgetting when retraining language models. Discover how it occurs, the signals, and practical methods to prevent loss.
• Chain-of-thought: Letting a model reason in stepsDiscover how to increase the reliability of AI models with Chain-of-Thought (CoT) prompting. Learn the theory, pitfalls, and practice with real-world…
• Chunking strategies for RAG: splitting documentsLearn how to effectively split documents for RAG. Compare fixed windows, recursive splitters, semantic chunking, and hierarchical document structures.
• Constitutional AI and RLAIF: training with AI feedbackLearn how Constitutional AI and RLAIF train language models using automated AI feedback and principles instead of manual human annotations.
• Context Engineering Explained: Why the Prompt is Not EnoughDiscover what context engineering is, how to assemble dynamic context (RAG, memory, tools), and why a larger context window is not always better.
• The token economy: what you pay per token and whyDiscover how the token economy of LLMs works, why input and output are priced differently, and how to get a grip on your API costs.
• Model Distillation Explained: How Small AI Models Learn fromDiscover how model distillation works: the technique that allows small AI models to adopt the knowledge of large models for faster and cheaper performance.
• DPO versus RLHF: two routes to preference | llmnet.nlA substantive comparison between RLHF and DPO for aligning language models with human preferences, how they work, and practical trade-offs.
• Learning path for AI agents: from prompt to systemDiscover the complete learning path for AI agents: from individual prompts to deterministic systems with state machines, tool calling, and robust validation.
• Embeddings explained: how machines understand meaningDiscover how AI understands text through embeddings and vector spaces. Learn the basics of similarity, semantic search, and clustering in this compact module.
• Evals for agents: why AI agents fail without testingDiscover how to systematically test AI agents with agent evals. Prevent infinite loops, incorrect API calls, and unpredictable behavior in production.
• Fine-tuning vs. prompting: when should you start training?Discover when to choose prompt engineering and when fine-tuning is necessary. Analyze costs, data requirements, latency, and model behavior in production.
• Fine-tuning vs prompting vs RAG: when to choose what?Discover when to choose prompting, RAG, or fine-tuning. Compare concepts, costs, complexity, and use the decision tree for your AI application.
• FlashAttention dissected: faster GPU computationDiscover how FlashAttention solves the quadratic memory bottleneck of transformers through GPU memory hierarchy, tiling, and online softmax.
• Memory and context: the difference in language modelsDiscover the essential difference between memory and context in language models. Learn how applications keep conversations coherent and what goes wrong.
• Memory in LLM apps: what does the model remember?Discover how memory works in LLM apps. Learn the difference between stateless inference and external memory layers such as vector stores and semantic buffers.
• Memory in LLM Applications: How AI Learns to Remember andDiscover how memory works in LLM applications. Learn the difference between context windows, conversation history, summaries, and how to forget in a…
• Grouped-query attention and memory usage explainedA detailed analysis of grouped-query attention (GQA). Learn how this technique reduces the memory usage of KV caching and LLM's versnelt.
• AI Guardrails Explained: The Safety Filters of LanguageDiscover how AI guardrails work, the difference between training alignment and runtime filters, and why language models sometimes refuse to answer.
• Why Does AI Sometimes Make Things Up? Understanding andDiscover why language models hallucinate, how to recognize incorrect AI answers, and which techniques (such as RAG) you can use to prevent this.
• How to assess whether a model answers well: evalsDiscover how to measure the quality of LLM responses with automated evaluations, test sets, deterministic checks, and LLM-as-a-judge methods.
• How an LLM works: the 10-minute basics courseDiscover in 10 minutes how a Large Language Model works. From tokenization to weights and probabilities: the complete basics clearly explained.
• How does an AI learn? Training explained for beginnersDiscover intuitively how artificial intelligence learns. Learn about data quality, pattern recognition, and the difference between training, fine-tuning, and…
• Evaluating AI responses: test sets and LLM-as-judgeDiscover how to measure the quality of AI responses with representative test sets, automated metrics, and LLM-as-judge evaluations.
• In-context learning: learning without weight updatesHow does an LLM learn purely based on examples in the prompt? Discover the mechanisms of in-context learning, induction heads, and the limits.
• AI Inference Explained: What Happens Under the Hood?Discover what AI inference is. Learn step-by-step how a language model processes your prompt, from tokens and the forward pass to generating a response.
• RAG explained: how to make your own documents searchable with AIRetrieval-Augmented Generation (RAG) lets an AI search your own documents before it answers. How it works step by step, plus a hands-on exercise to design your own pipeline.
• How KV caching is built up in transformer architecturesLearn how KV caching works in transformer architectures, how the cache is built up in VRAM and how GQA and MQA reduce memory usage.
• Quantization explained: INT4, FP8 and model compressionDiscover how quantization works in LLMs. Learn the difference between INT4, FP8, PTQ, and QAT to optimize memory use and compute time effectively.
• Getting Started with LLMs: From Zero to Your FirstDiscover how Large Language Models (LLMs) work in this practical learning path. Learn effective prompting, API basics, and key privacy guidelines.
• LoRA and Adapters Explained: Cheap Fine-Tuning | llmnet.nlDiscover how LoRA and adapters make fine-tuning LLMs affordable and accessible. Learn the technique, benefits, and implementation for efficient AI.
• Machine unlearning: erasing data from network weightsHow do you erase data from a trained neural network without full retraining? Discover techniques, mathematics, and verification methods for machine unlearning.
• Mechanistic interpretability explained - llmnet.nlA thorough explanation of mechanistic interpretability in AI. Discover how circuits, superposition, and the residual stream work.
• Mixture-of-Experts (MoE) ExplainedDiscover how Mixture-of-Experts (MoE) makes modern AI models faster and smarter by not using all parameters at once. Read how it works.
• Model-specific prompting: Claude, Gemini & ChatGPTDiscover why Claude, Gemini, and ChatGPT each require a unique approach. Optimize your prompts for system cards, context phenomena, and architecture.
• Multimodal input: text, image, and audio in one modelDiscover how multimodal LLMs process text, images, and audio in a single transformer. From patch projection to cross-modal representations and compute costs.
• Multimodal tokenization of images and audio explainedHow do multimodal AI models convert images and audio into tokens? Discover the mechanisms of patch embedding, audio codecs, and context space.
• Parameters and weights: what is in an AI model? | LearnDiscover what parameters and weights in an AI model are, why size matters, and the difference between training and inference with a clear analogy.
• Positional Encoding Explained: Order in TransformersHow do transformer models understand word order? Discover how sinusoidal, learned, and rotary positional encodings like RoPE work.
• Prompt injection and jailbreaks: why instructions and data get mixed upPrompt injection and jailbreaks explained: why an LLM can't separate instructions from data, how RAG increases the risk, and what you can do about it.
• Prompting for Non-Technical UsersDiscover how to write smart AI prompts as a non-technical user. Learn to ask clear questions, provide context, and iterate for immediate results without code.
• Prompt guide per model: Claude, Gemini, and ChatGPTDiscover how to prompt Anthropic Claude, Google Gemini, and OpenAI ChatGPT differently with specific prompting techniques for each AI model.
• RAG Explained for Beginners: Making Your Own DocumentsLearn the basics of Retrieval-Augmented Generation (RAG). Discover how embeddings and vector databases work and make your own data searchable with AI.
• Re-ranking explained: cross-encoders vs vector similarityDiscover how re-ranking with cross-encoders improves search tasks in RAG compared to bi-encoders and cosine similarity in terms of latency and precision.
• ReAct patterns dissected: reasoning and actionDissect the ReAct pattern for LLM agents: discover how Thought, Action, and Observation come together to combine reasoning with external tools.
• Reasoning models: what changes when thinking becomes a separate stepReasoning models explained: what changes when thinking becomes a separate step, how test-time compute works, and when to use a reasoning model.
• Reinforcement learning - the basics explainedDiscover how reinforcement learning works through reward and punishment. View the core concepts, examples, and practical applications.
• What is RLHF? Reinforcement Learning from Human Feedback andDiscover how AI models are aligned using RLHF and DPO. Learn how human feedback ensures safe, helpful, and nuanced responses from LLMs.
• Rotary Position Embedding and the processing of long textsRotary Position Embedding (RoPE) encodes token positions as relative angles. Read how it works and why long texts benefit from it.
• Temperature, top-p and other sampling parameters explainedDiscover how temperature, top-p, and penalties affect the output of LLMs. Learn when to choose creativity or reliability with practical examples.
• Scaling laws: compute and data sizeHow do scaling laws predict language model performance? Learn all about compute, data size, Chinchilla ratios and data walls.
• Self-attention dissected mathematically: dot product to softmaxA mathematical dissection of scaled dot-product attention in transformers: from query, key, and value projections to softmax and matrix multiplication.
• Speculative decoding in LLMs explained | llmnet.nlRead how speculative decoding increases language model generation speed without quality loss through parallel token verification.
• Synthetic data: how models learn from AI dataDiscover how AI models train on synthetic data, what the risks of model collapse are, and how automated validation safeguards quality.
• System prompts dissected: instructions and priorityHow do system prompts get priority in LLMs? Read about chat templates, role tags, attention mechanisms, and the limits of steering a model.
• Temperature and Top-p Practically ExplainedDiscover how temperature and top-p work in LLMs. Practical examples, mathematical explanation, and recommended settings for every AI application.
• Tokenization explained: how an LLM chops text into piecesDiscover what tokenization is, how large language models process text, and why this is crucial for API costs and the context window of your AI application.
• Tool calling and structured output explainedDiscover how an LLM calls functions and generates structured JSON. Comprehensive guide about JSON Schema, grammar constraints, and tool calling.
• Interactive KV Cache Memory CalculatorCalculate exactly how much VRAM your LLM's Key-Value cache uses based on context length, layers, hidden size, GQA, and precision.
• Visual tokenizer simulator - how LLM text processing worksDiscover how a large language model breaks text into smaller pieces or tokens with this interactive and visual simulator for Dutch and English texts.
• Training Data and Bias in Language ModelsDiscover how bias in AI arises from training data, the impact of English on Dutch output, and practical methods to mitigate bias.
• Tree-of-Thoughts and Graph-Reasoning ExplainedDiscover how Tree-of-Thoughts and Graph-of-Thoughts add search trees, branching, and backtracking to language models for complex planning and reasoning.
• From Text to Numbers: Vector Embeddings in Practice | LLMNet LearnDiscover how vector embeddings convert text into numbers for AI. Practical explanation of dimensions, cosine similarity, and RAG applications in search systems.
• Vector Indexing with HNSW and IVF Explained | llmnetA thorough explanation of vector indexing with HNSW and IVF. Discover how approximate search, recall, settings, and compression affect search performance.
• Forward vs Backward Pass: BackpropagationDiscover the mathematical difference between the forward and backward pass in transformers, including memory usage and gradient computation.
• What is the Difference Between an Agent and a Chatbot? | leren.llmnet.nlDiscover the difference between an AI chatbot and an AI agent. Read how the loop principle, tool use, and autonomous steps work without complicated jargon.
• What is a transformer? The brain behind modern AIDiscover how the transformer model works behind modern AI. Learn about the attention mechanism, simple analogies, and test your knowledge with a practical…
• What does a token cost? Context length, input, and outputCalculate what a token costs: input versus output, context length, and when caching reduces the bill. With practical examples and decision rules.