AI & Tech Glossary

Software that pursues a goal by taking a sequence of actions on its own, rather than responding once to a single prompt. An agent decides what to do next based on the result of its last action, calling tools, reading files, or browsing the web as needed until the task is done or it needs your input.

The broader shift toward AI systems that act autonomously across multiple steps, as opposed to "generative AI" that produces a single output per request. Agentic AI is the category; an AI agent is one instance of it.

An open standard that lets AI models connect to external tools, files, and data sources in a consistent way, instead of every app building its own custom integration. Most 2026 coding and business agents use MCP under the hood to reach your codebase, CRM, or file system.

Writing software by describing what you want in plain language and running whatever an AI generates, without reading the code line by line. The defining trait isn't using AI to code, it's choosing not to review the output before it ships.

Optimizing content to be cited by AI systems like ChatGPT, Perplexity, and Google AI Overviews, rather than (or in addition to) ranking in traditional blue-link search results. Overlaps heavily with good SEO, plus a few AI-specific technical items like crawler access and llms.txt.

A plain-text file at a website's root that gives AI crawlers a structured, machine-readable index of the site's content, similar in spirit to a sitemap but written for models instead of search engine crawlers. Not a confirmed ranking factor, but cheap to implement.

Google's AI-generated summary shown above traditional search results for many queries. Overviews draw their citations overwhelmingly from pages that already rank well organically, so classic SEO remains the foundation even for AI-citation visibility.

A technique where an AI model looks up relevant information from an external source (documents, a database, the web) before generating its answer, instead of relying only on what it learned during training. This is how tools like NotebookLM and most "chat with your docs" products work.

The class of AI model (GPT, Claude, Gemini, and similar) trained on large amounts of text to predict and generate language. Almost every AI tool covered on this site, from chatbots to coding assistants, is built on top of an LLM.

The amount of text (measured in tokens) an AI model can "see" and reason about at once, including your prompt, any attached files, and its own prior responses in the conversation. A larger context window means the model can work with bigger codebases or longer documents without losing track of earlier details.

When an AI model states something false or fabricated with the same confidence as a true statement. The main reason to fact-check AI output before publishing or acting on it, especially for anything specific: statistics, citations, prices, or version numbers.

Additional training applied to an existing AI model on a narrower, specific dataset to specialize its behavior, as opposed to prompting the general-purpose model as-is. Increasingly less common for everyday use now that long context windows and RAG let you get specific results without retraining a model.

The practice of structuring instructions to an AI model to reliably get the output you want: providing examples, specifying format, breaking a task into steps. Matters less than it used to as models have gotten better at inferring intent, but still meaningfully affects output quality for complex or ambiguous tasks.

AI processing that happens locally on a phone or computer rather than being sent to a remote server. Faster and more private, but limited by the device's hardware compared to a cloud model. Apple Intelligence and most on-phone camera/voice features use a mix of on-device and cloud processing depending on the task.

A password replacement based on public-key cryptography, tied to your device and unlocked with a fingerprint, face scan, or PIN instead of a typed password. Resistant to phishing and credential-stuffing attacks in a way traditional passwords aren't, since there's no shared secret to steal.

Connecting apps and automating multi-step workflows (an order coming in, a form being submitted, a file being uploaded) using a visual builder instead of writing code. Zapier, Make, and n8n are the three most common platforms.

A Samsung security error that appears during recovery mode or firmware flashing when the device detects a mismatch between the bootloader, firmware, or partition data and what it expects. Model-specific and version-specific; there's no universal fix.

A background system process on both iPhone and Samsung devices tied to device-to-device continuity features (things like proximity-based handoff between your devices). Frequently shows up in battery-usage stats without explanation, which is why it gets searched so often.

Samsung's Android skin layered on top of stock Android, covering the interface, default apps, and Samsung-specific features across Galaxy phones and tablets. Version numbers (One UI 7, 8, 8.5) roughly track the underlying Android version but ship on Samsung's own schedule.

Apple's translucent, light-refracting interface design introduced in iOS 26, applied across system UI elements like the control center, navigation bars, and app icons. Divisive enough that a meaningful share of users search for how to turn it off.