Tech terms, inplain English.
Short, clear explanations of terms people encounter in technology, business, work, and culture.
30-sec reads. No ads. No login.
Tuesday, November 3
Autonomous AI
AI & Automation
The dictionary says
In short
Set the goal, step back.
It keeps working without asking you about every step.
Browse all terms
86 terms
Activation Steering
A technique that changes an AI model's behavior by modifying internal numerical signals while the model is processing information.
Agentic AI
AI systems that can independently pursue goals, make decisions, and take actions in dynamic environments, often with minimal human oversight.
Agile
An approach to software development that emphasizes collaboration, frequent delivery of working software, and adapting plans through feedback.
AI Agent
An AI system that can pursue a goal by deciding what actions to take, using available tools or information, and responding to the results of those actions.
AI Evals
Structured tests used to measure how accurately, reliably, or effectively an AI model or system performs specific tasks.
API
Application programming interface: a defined interface that allows software systems to request data or functionality from one another.
ARR
Annual recurring revenue: the recurring subscription revenue a business expects from active customer contracts normalized to a one-year period.
Autonomous AI
AI systems designed to operate with limited ongoing human direction by making decisions and taking actions toward defined goals.
Bandwidth
The maximum amount of data a connection can carry in a given amount of time, usually measured in megabits or gigabits per second.
Cache
A temporary storage layer that keeps previously retrieved or computed data so future requests can be served faster.
CDN
Content delivery network: geographically distributed network of servers that caches and delivers content from locations closer to users.
Chain of Thought
The step-by-step reasoning a model writes out before giving its final answer.
Churn
The loss of customers, subscribers, or recurring revenue over a given period.
CI/CD
Continuous integration and continuous delivery or deployment: practices for frequently combining software changes and automating their testing, release preparation, and, in some workflows, deployment.
Cloud Computing
On-demand access to shared computing resources such as servers, storage, networking, and applications over a network.
Computer Use
An AI capability that lets a model operate a computer like a person: looking at the screen, moving the pointer, clicking and typing.
Container
An isolated software environment that packages an application with the dependencies it needs to run consistently.
Context Engineering
Designing what information an AI model receives with each request, such as instructions, documents, tool results and memory, so it has what it needs to do the job.
Context Window
The amount of information a language model can consider at one time when processing an input and generating a response.
Credit
A unit of allotted or prepaid usage that an AI service deducts when someone uses certain features.
Database
An organized collection of data managed so applications can store, retrieve, update, and query information.
DeepInfer
An open-source inference framework from DeepSeek for running DeepSeek models.
DevOps
A way of developing and running software that combines shared responsibility, collaboration, and automation to improve delivery and reliability.
Distillation
A training method where a smaller "student" model learns to copy the answers of a larger "teacher" model.
DNS
Domain Name System: a distributed naming system that connects domain names with records such as IP addresses and mail-server information.
Embeddings
Numerical representations that encode the meaning or features of data so computers can compare items by similarity.
Encryption
The cryptographic process of transforming readable data into ciphertext so it cannot be understood without the appropriate means to decrypt it.
ETL
Extract, transform, load: a data integration process that extracts data from source systems, transforms it into a useful form, and loads it into a target system.
Fine-Tuning
Additional training applied to an existing model using a narrower dataset so the model becomes better suited to a specific task or behavior.
Generative AI
AI that produces content such as text, images, audio, video, or code using patterns learned during training.
Golden Query
A reviewed test query paired with expected relevant results, used to evaluate a search or AI retrieval system.
Grounding
Connecting an AI model's responses to relevant external information so its answers can be more accurate and better supported.
Guardrails
Rules and checks placed around an AI system to keep its inputs and outputs safe, on-topic and within policy.
Hallucination
An AI output that presents incorrect, unsupported, or invented information as though it were reliable.
HTTP
Hypertext Transfer Protocol: the set of rules clients and servers use to exchange requests and responses on the web.
Incubator
A program that helps very early-stage startups get going with workspace, mentoring and shared resources, often over several months or longer.
Inference
The process of using a trained AI model to generate predictions, answers, or other outputs from new inputs.
Inference-Time Compute
The computing resources an AI model uses to process a request and generate an answer after it has been trained.
Jailbreak
A prompt or trick designed to get an AI model to ignore its safety rules and produce something it's meant to refuse.
JSON
JavaScript Object Notation: a text format for representing structured data that software systems can store and exchange.
KPI
Key performance indicator: a measure chosen to track progress toward an important business or service objective.
Kubernetes
An open-source platform for deploying, managing, and scaling containerized applications across clusters of computers.
Latency
The delay between sending a request and getting a response, usually measured in milliseconds.
LLM
A large language model: an AI model trained on large amounts of text to understand and generate language.
Load Balancer
A system that distributes incoming network traffic across multiple servers or application instances.
Machine Learning
An approach to AI in which models learn patterns from data and use them to make predictions or produce other outputs.
MCP
Model Context Protocol: an open standard for connecting AI applications to external tools and data sources.
MFA
Multi-factor authentication: an authentication method that requires more than one distinct type of factor before access is granted.
Mixture of Experts
A way of building AI models from many smaller specialist parts, where only a few are switched on for each piece of input.
Moat
A durable competitive advantage that helps protect a business from rivals.
Model Steering
Guiding an AI model's responses or behavior toward a desired outcome using instructions, settings, or techniques that influence how it operates.
MRR
Monthly recurring revenue: the recurring subscription revenue a business expects from active customer contracts normalized to a one-month period.
Multi-Agent System
A setup where several AI agents, each with its own role, work together on a task, often with one agent coordinating the others.
Multimodal AI
AI that can work with more than one type of data, such as text, images, audio, or video.
MVP
Minimum viable product: an early version of a product used to test important assumptions about customers with the least effort needed to learn.
OAuth
An authorization framework that allows an application to receive limited access to another service without requiring the user to share their password with that application.
Object Storage
A storage model that keeps data as individually addressable objects with metadata rather than organizing it primarily as files in folders or blocks on a disk.
Open-Weight Model
An AI model whose trained weights are published, so anyone can download it, run it on their own hardware and adapt it.
Prompt
Input supplied to an AI model, such as a question, instruction, or example, to guide its response.
Prompt Injection
An attack that hides instructions in text an AI reads, tricking it into ignoring its rules or doing something its user never asked for.
RAG
Retrieval-augmented generation: a technique that gives an AI model access to external information at request time so it can generate responses using retrieved context.
Reasoning Model
An AI model designed to use additional computation to work through complex problems before producing an answer.
Red Teaming
Deliberately attacking your own system, the way a real adversary would, to find weaknesses before someone else does.
REST API
An Application Programming Interface designed around resources and the architectural constraints of Representational State Transfer, commonly using HTTP methods and URLs.
Runway
How many months a company can keep operating before it runs out of cash at its current rate of spending.
SaaS
Software as a service: a cloud service model in which customers use software applications operated by a provider rather than managing the underlying infrastructure themselves.
Sandbox
An isolated space where code, apps or AI agents can run and be tested without touching real systems or data.
SDK
Software development kit: collection of tools, libraries, documentation, and utilities that helps developers build software for a particular platform, service, or environment.
Semantic Cache
A cache that reuses stored results for inputs with similar meaning rather than requiring exact matches.
Serverless
A cloud application model in which the provider manages much of the underlying server infrastructure, scaling, and runtime operations for the developer.
Small Language Model
A language model built to be much smaller than the largest models, so it runs faster and cheaper, often directly on a phone or laptop.
SQL
Structured Query Language: a language for defining, retrieving, and changing data in relational databases.
SSO
Single sign-on: an authentication approach that lets one account and sign-in be used to access multiple applications.
Steering Vector
A numerical direction used to influence an AI model's internal representations and encourage a particular behavior.
Sycophancy
An AI model's tendency to agree with a user's views or assumptions at the expense of accuracy.
Synthetic Data
Data created by a program, often an AI model, instead of collected from real people or events, used to train or test other systems.
Temperature
A model-generation setting that adjusts the randomness of token selection, influencing how varied an AI model’s output can be.
Token
A unit of text, such as a word, word fragment, or punctuation mark, that a language model processes or generates.
Tool Calling
A capability that lets an AI model request the use of an external tool, function, API, or service as part of completing a task.
Vector Database
A database designed to store, index, and search vector embeddings so similar items can be retrieved efficiently.
Verbosity
The amount of detail and length in an AI model's response.
Version Control
A system for recording changes to files over time so people can compare revisions, recover earlier versions, and coordinate work.
Vibe Coding
Building software by describing what you want to an AI and accepting the code it writes, often without reading it closely.
Virtual Machine
A software-based computer that runs its own operating system using virtualized computing resources.
Webhook
An event-driven HTTP notification that one system sends to another when a specific event occurs.
Zero Trust
A security approach that continually evaluates access based on identity, device, context, and policy rather than assuming users or systems are trustworthy because of their network location.