Open AI glossary
NAiOS Wiki
A plain-language dictionary of AI terminology — concepts, techniques, architectures and trends — explained clearly and kept up to date. 73 terms.
An educational, independent glossary. Some terms are implemented by NAiOS; others — such as risks or bad practices (prompt injection, deepfakes) — are included only to understand them, not as platform features.
TrendingTOPS (Tera Operations Per Second)MultimodalNPU (Neural Processing Unit)World ModelsInference Time ComputeHarness
| Term | Category | Definition |
|---|---|---|
| AGI (Artificial General Intelligence) | Tendencias y Cultura | Artificial General Intelligence, capable of any human intellectual task |
| AI Act (EU AI Act) | Gobernanza y Estrategia | Regulation (EU) 2024/1689: the first comprehensive law on artificial intelligence, with obligations based on risk level |
| AI Agents | Agentes y Autonomía | Systems that execute multi-step tasks without constant supervision |
| AI Alignment | Seguridad y Ética | Aligning AI goals with human values |
| AI Brain Fry | Tendencias y Cultura | Acute cognitive fatigue from excessive use or supervision of AI tools. Term coined by researchers from BCG and the University of California (Riverside) in 2026. |
| AI Copilot | Agentes y Autonomía | AI Assistant integrated into work tools |
| AI Governance | Gobernanza y Estrategia | Regulatory and management framework for the responsible use of AI |
| AI Overview | Tendencias y Cultura | AI-generated summaries in search results |
| AI Psychosis | Tendencias y Cultura | Informal term (not a recognized clinical diagnosis) for psychotic-like symptoms—delusions, paranoia—associated with the intensive use of AI chatbots. |
| AI Safety | Seguridad y Ética | Discipline focused on ensuring that AI is safe |
| AI Wrapper | Gobernanza y Estrategia | Application that simply wraps an existing model without adding real value |
| Agentic AI | Agentes y Autonomía | AI capable of planning, deciding, and acting autonomously |
| Agentic Workflows | Agentes y Autonomía | Workflows where AI agents act autonomously |
| Algorithmic Bias | Seguridad y Ética | When AI reproduces or amplifies biases present in the data |
| Chain of Thought (CoT) | Razonamiento y Técnicas de Prompting | Technique where AI thinks step-by-step before responding |
| Coding agent | Agentes y Autonomía | An AI agent that writes, modifies and tests code on its own within a project |
| Context Window | Razonamiento y Técnicas de Prompting | Amount of text a model can remember in a conversation |
| Data Contamination | Entrenamiento y Datos | When evaluation data leaks into training, biasing benchmarks |
| Data Moat | Entrenamiento y Datos | Competitive advantage of a company thanks to its unique data |
| Decision Intelligence | Gobernanza y Estrategia | Using AI to improve business decision-making |
| Deepfake | Generación de Contenido | Hyper-realistic AI-generated fake content |
| Edge AI | Infraestructura y Despliegue | Run AI directly on the device, without relying on the cloud |
| Embeddings | Entrenamiento y Datos | Numerical representation of the meaning of texts or images |
| Fine-Tuning | Entrenamiento y Datos | Retraining a model with specific data for a use case |
| Frontier Models | Arquitectura y Modelos | The most advanced and powerful models at all times |
| GEO (Generative Engine Optimization) | Tendencias y Cultura | The new SEO, optimizing content to be cited by generative AI |
| Generative AI | Generación de Contenido | Artificial intelligence that creates new content (text, image, video, audio) |
| Grounding | Entrenamiento y Datos | Grounding AI responses in real and verifiable data |
| Guardrails | Seguridad y Ética | Restrictions and limits imposed on AI to prevent harm |
| Hallucination | Seguridad y Ética | When AI generates false information with total confidence |
| Harness | Agentes y Autonomía | The infrastructure around the model that turns it into an agent: loop, tools, context, permissions and traces |
| Human in the Loop (HITL) | Agentes y Autonomía | A design in which a person reviews or approves the relevant actions of an AI system before they take effect |
| Human oversight | Gobernanza y Estrategia | An obligation under the EU AI Act for trained individuals to be able to understand, monitor, and intervene in an AI system |
| Inference Time Compute | Razonamiento y Técnicas de Prompting | Dedicate more compute capacity during the response, not training |
| Jailbreaking | Seguridad y Ética | Techniques for evading model security restrictions |
| LLM (Large Language Model) | Arquitectura y Modelos | Large language model, the foundation of today's chatbots |
| LoRA / QLoRA | Entrenamiento y Datos | Techniques for adapting models with limited resources |
| Local Inference | Infraestructura y Despliegue | Running AI models on your own hardware |
| Loop Engineering | Generación de Contenido | Designing the autonomous loops that guide code agents (2026 term) |
| Mixture of Experts (MoE) | Arquitectura y Modelos | Architecture that activates only parts of the model based on the task |
| Model Context Protocol (MCP) | Infraestructura y Despliegue | Standard protocol for models to connect with external tools |
| Model Distillation | Entrenamiento y Datos | Compressing a large model into a small one while preserving quality |
| Model Welfare | Seguridad y Ética | Ethical debate on whether advanced models deserve some form of consideration |
| ModelOps | Gobernanza y Estrategia | Operational management of the AI model lifecycle |
| Multi-Agent Orchestration | Agentes y Autonomía | Coordinate multiple AI agents to solve complex tasks together |
| Multimodal | Arquitectura y Modelos | Models that understand and generate text, image, audio and video |
| NPU (Neural Processing Unit) | Infraestructura y Despliegue | A chip specialized in accelerating neural network inference with maximum energy efficiency |
| No-Code AI | Tendencias y Cultura | AI tools that require no programming to use |
| Prompt Drift | Razonamiento y Técnicas de Prompting | When a prompt that worked stops delivering good results over time |
| Prompt Engineering | Razonamiento y Técnicas de Prompting | Design and optimization of prompts for better AI results |
| Prompt Injection | Seguridad y Ética | Attacks that manipulate AI by inserting malicious instructions |
| RAG (Retrieval Augmented Generation) | Entrenamiento y Datos | Combine real data search with text generation |
| RLHF | Entrenamiento y Datos | Reinforcement Learning from Human Feedback |
| Reasoning Models | Razonamiento y Técnicas de Prompting | Models designed specifically to reason, not just predict text |
| Red Teaming | Seguridad y Ética | Testing an AI by attempting to break it to find vulnerabilities |
| Scaffold | Agentes y Autonomía | The code structure that connects and orchestrates an AI agent |
| Shadow AI | Gobernanza y Estrategia | Unauthorized use of AI tools within organizations |
| Small Language Models (SLM) | Arquitectura y Modelos | Lightweight language models, designed to run on devices |
| Sovereign AI | Gobernanza y Estrategia | Strategic priority for countries to have their own AI infrastructures |
| Superintelligence | Tendencias y Cultura | AI that surpasses human intellectual capacity in all fields |
| Synthetic Data | Entrenamiento y Datos | Artificially generated data for training models |
| Synthetic Media | Generación de Contenido | Media (image, audio, video) artificially created by AI |
| TOPS (Tera Operations Per Second) | Infraestructura y Despliegue | Unit that measures the raw performance of an AI accelerator: trillions of operations per second |
| Temperature | Razonamiento y Técnicas de Prompting | Parameter that controls AI creativity vs. precision |
| Text-to-Video | Generación de Contenido | Generate video from a text description |
| The Great Integration | Tendencias y Cultura | Informative label for the phase of mass and cross-sector AI adoption across all industries. It is not an academic term or a quote from MIT. |
| Tokenization | Entrenamiento y Datos | Process of dividing text into units (tokens) that the model can process |
| Tokenmaxxing | Razonamiento y Técnicas de Prompting | Maximizing a model's context window. Colloquially, also: using many tokens as a (debatable) productivity metric. |
| Transformer | Arquitectura y Modelos | The base architecture of almost all current language models |
| Vibe Coding | Generación de Contenido | Program by describing what you want and letting the AI write the code |
| Voice Cloning | Generación de Contenido | Clone a human voice with AI from a few seconds of audio |
| World Models | Arquitectura y Modelos | Models that learn the dynamics of an environment, not just text |
| Zero-Shot | Razonamiento y Técnicas de Prompting | Ability of a model to perform tasks without being specifically trained |
From the NAiOS blog
The latest on applied AI.
NAIOS FUNCTIONS
Support in 2026: an entire department, not a mailbox
5 de octubre de 2026
NAIOS FUNCTIONS
NAiOS, explained today: the AI chat that already includes what your business pays for separately
4 de octubre de 2026
NAIOS FUNCTIONS
Skills: instructions that the chat applies on its own
3 de octubre de 2026