Row 1 · Primitives
Atomic building blocks that everything else in AI is composed from.
PR
Prompt
Instructions given to an AI.
Reactive
EM
Embeddings
Numerical representations of meaning.
Retrieval
LLM
Large Language Model
Foundational models like ChatGPT, Claude, Granite.
Models
Row 2 · Compositions
Elements formed by combining primitives into higher-level capabilities.
FC
Function Calling
LLMs invoking tools or APIs before answering.
Reactive
VX
Vector Databases
Semantic search over embeddings.
Retrieval
RAG
Retrieval Augmented Generation
Retrieved context augments LLM prompts.
Orchestration
GR
Guardrails
Runtime safety and output validation.
Validation
M
Multi‑modal Models
Text, image, audio capable models.
Models
Row 3 · Deployment
Elements focused on putting AI systems into production.
AG
Agents
Think–act–observe loops to reach goals.
Reactive
FT
Fine‑tuning
Adapting models with domain‑specific data.
Retrieval
FW
Frameworks
Platforms like LangChain & LangGraph.
Orchestration
RT
Red Teaming
Adversarial testing and jailbreak discovery.
Validation
SM
Small Models
Fast, cheap, distilled or edge‑ready models.
Models
Row 4 · Emerging
Rapidly evolving concepts pushing the frontier of AI.
MA
Multi‑agent Systems
Multiple AIs collaborating and coordinating.
Reactive
SY
Synthetic Data
AI‑generated training data.
Retrieval
IN
Interpretability
Understanding why models behave as they do.
Validation
TH
Thinking Models
Models that reason before answering.
Models