A practical knowledge base for AI protocols, agents, core concepts, frameworks, and tools.
A reasoning model is a large language model that produces a longer thinking process before it answers. By 2026, whether to think and for how long has become an adjustable reasoning effort setting on mainstream models, and the setting you pick directly affects quality, latency, and cost.
New words arrive fast in AI. This series explains common AI terms like LLM, Transformer, RAG, Agent, MCP, agent memory, and embodied AI in everyday language, with no programming or math background required, helping you turn scattered jargon into a coherent, reusable mental map.
Vibe Coding is a term coined by Andrej Karpathy in 2025, describing a way of programming where you hand many code details over to AI and mainly describe your intent in natural language, run the results, and keep adjusting.
In AI, distillation usually refers to "knowledge distillation" or "model distillation." Its core idea is to use a larger, more capable teacher model to guide a smaller, lighter student model, so the student model learns the teacher model's behavior as much as possible.
Reinforcement Learning, often abbreviated RL, is a class of machine learning methods. What it focuses on isn't "giving the model a pile of correct answers to learn from," but letting an agent try actions in an environment and gradually learn a better strategy based on rewards and penalties.
In Edge AI, "edge" doesn't mean being marginalized—it refers to devices and network locations close to where data is generated.
GEO stands for Generative Engine Optimization. It's a new term that emerged after SEO, focused not on where a web page ranks in traditional search results, but on whether your content, brand, or viewpoint can be cited, understood, and recommended by AI search, chat assistants, and generative answers.
In an AI context, fine-tuning refers to continuing to train an existing base model with additional data, so the model better fits a certain kind of task, format, style, or domain.
An AI workflow is a process that strings together multiple AI steps, tools, and human-confirmation points. It doesn't have to be complex—it can start with a small flow like "receive an email, automatically summarize the content, generate a draft reply, and have a person confirm before sending."
As the number of available models grows, one question becomes obvious: which model should I use? Some models are good at reasoning, some are cheap, some are fast, some support images, some have a long context window, and some are more stable in certain regions or at certain times. Model routing exists to serve this question.
An AI gateway is common infrastructure when enterprises and teams use multiple models. It usually sits between applications and models, uniformly managing access, permissions, cost, security, logging, and routing.
Inference is easily misunderstood as referring only to complex logical reasoning. In fact, in a machine learning context it's broader: it refers to using an already-trained model to produce predictions, classifications, generations, or decisions for new inputs.
Showing 12 of 12 entries
Every entry grouped by topic — one click straight to the explanation.
