Technical Level

What it is: Techniques for creating prompts with clear, hierarchical structure.

Why: Well-structured prompts produce more accurate and consistent responses.

Concepts: Headers, sections, XML/Markdown delimiters, instruction hierarchy.

What it is: Organize prompts into distinct blocks for each type of information.

Why: Avoids confusion and improves the model's interpretation.

Concepts: System blocks, context, input data, specific instructions.

What it is: Clearly define the goal before drafting the prompt.

Why: Focusing on the desired outcome increases effectiveness.

Concepts: SMART goals, success criteria, quality metrics.

What it is: Select and position examples for maximum impact.

Why: Well-chosen examples establish patterns and expectations.

Concepts: Diversity, representation, ordering, ideal quantity.

What it is: Eliminate multiple interpretations in prompts.

Why: Ambiguity is the main cause of incorrect answers.

Concepts: Explicit definitions, technical terms, limited scope.

What it is: Create parameterizable prompt structures.

Why: Saves time and ensures consistency.

Concepts: Placeholders, variables, dynamic templates, prompt libraries.

What it is: Diagnose and fix prompts that don't work.

Why: Systematic debugging speeds up optimization.

Concepts: Error patterns, A/B testing, systematic iteration.

What it is: Explicitly specify what NOT to do.

Why: Avoids common unwanted behaviors.

Concepts: Explicit prohibitions, behavior limits, guardrails.

What it is: Set clear limits for length, format, and scope.

Why: Controls verbosity and keeps the focus.

Concepts: Word limits, maximum bullets, controlled depth.

What it is: Control the structure, hierarchy, and presentation of the output.

Why: Structured outputs are more useful and parseable.

Concepts: JSON, XML, Markdown, tables, schemas.

What it is: Start the model's response with predefined text.

Why: Ensures the format and guides the style.

Concepts: Assistant prefill, JSON starter, forced structure.

What it is: Adjust voice, formality, and technical level.

Why: Tailors the response to the target audience.

Concepts: Formality scales, expertise levels, style personas.

What it is: Ask the model to check its own response.

Why: Reduces errors and increases reliability.

Concepts: Self-check, format validation, fact-checking.

What it is: Create consistent patterns for outputs at scale.

Why: Facilitates integration and automated processing.

Concepts: Schemas, API contracts, automated validation.

What it is: Break complex problems into manageable subtasks.

Why: Improves accuracy and enables focused processing.

Concepts: Task breakdown, dependencies, parallelization.

What it is: Connect prompts in sequence, where the output of one becomes the input of the next.

Why: Enables complex, specialized workflows.

Concepts: Pipelines, transformations, aggregation.

What it is: Prompts that guide you through multiple phases.

Why: Structure complex thinking within a prompt.

Concepts: Sequential phases, checkpoints, branches.

What it is: 3-phase framework for complex tasks.

Why: Separates strategic thinking from execution.

Concepts: Planning prompts, execution prompts, review prompts.

What it is: Improve outputs through feedback cycles.

Why: Achieves higher quality through iteration.

Concepts: Feedback loops, improvement criteria, convergence.

What it is: Manage information across multiple prompts.

Why: Prevents information loss and maintains coherence.

Concepts: Context passing, summarization, state management.

What it is: Real examples of workflows with chained prompts.

Why: Learn by applying concepts to real-world scenarios.

Concepts: Case studies, workflow templates, benchmarks.

What it is: How models process text + image/audio/video.

Why: Understanding the model leads to more effective prompts.

Concepts: Vision encoders, cross-modal attention, grounding.

What it is: Prompt structure for DALL-E, Midjourney, Stable Diffusion.

Why: Each model has its own syntax and style.

Concepts: Subject, style, medium, lighting, composition.

What it is: Specify what NOT to include in the image.

Why: Avoids artifacts and unwanted elements.

Concepts: Negative prompts, exclusions, quality tags.

What it is: Maintain consistent style and characters across images.

Why: Essential for professional projects and branding.

Concepts: Style seeds, character sheets, reference images.

What it is: Prompts for Sora, Runway, Pika, and video generators.

Why: Video adds a temporal dimension to prompts.

Concepts: Motion, timing, transitions, camera movement.

What it is: Create visual narrative sequences with prompts.

Why: Visual planning for videos and animations.

Concepts: Scene breakdown, shot types, continuity.

What it is: Real-world multimodal use cases in production.

Why: See how companies use multimodal.

Concepts: Marketing, e-commerce, education, entertainment.

What it is: Persistent instructions that define baseline behavior.

Why: Establishes context and rules for every conversation.

Concepts: System vs. user prompts, priority, persistence.

What it is: Architecture and structure of effective system prompts.

Why: Well-designed system prompts are the foundation of good assistants.

Concepts: Sections, hierarchy, embedded examples, guardrails.

What it is: Define an identity and expertise that persist throughout the conversation.

Why: Creates specialized, consistent assistants.

Concepts: Character design, knowledge boundaries, voice consistency.

What it is: Prompts that enable autonomous actions and tool use.

Why: Agents are the future of AI automation.

Concepts: Tool use, function calling, action planning.

What it is: Limits and safeguards for model behavior.

Why: Security and compliance in production.

Concepts: Safety rules, content filters, scope limitations.

What it is: Manage information between interactions.

Why: Agents need to “remember” to be useful.

Concepts: Short/long-term memory, state persistence, summarization.

What it is: Coordinate multiple agents in complex systems.

Why: Multi-agent systems are more powerful.

Concepts: Agent routing, delegation, consensus, collaboration.

What it is: Strategic design of how information is organized in context.

Why: Maximizes use of the context window.

Concepts: Prioritization, semantic chunking, context stuffing.

What it is: Combining LLMs with knowledge base search.

Why: Answers based on up-to-date, specific data.

Concepts: Retrieval, augmentation, generation pipeline.

What it is: Vector representations of text for semantic search.

Why: Foundation for RAG and similarity search.

Concepts: Embedding models, similarity metrics, dimensionality.

What it is: Databases optimized for vector search.

Why: Essential infrastructure for RAG at scale.

Concepts: Pinecone, Weaviate, ChromaDB, Qdrant, indexing.

What it is: Strategies for handling very long documents.

Why: Not everything fits in the context window.

Concepts: Chunking strategies, overlap, semantic splitting.

What it is: Methodologies for testing and evaluating prompt quality.

Why: Ensure performance before production.

Concepts: Eval frameworks, A/B testing, benchmarks, metrics.

What it is: Prepare prompts for large-scale deployment.

Why: Production has requirements for cost, latency, and reliability.

Concepts: Caching, batching, model selection, cost optimization.