Knowledge index

Code organized around a clear responsibility, implemented as functions, a file, or several files, collaborating through defined interfaces.

How closely work inside a module belongs together. High cohesion groups work serving a focused responsibility and related reasons for change.

The degree of dependency between modules. Low coupling reduces reliance on another module's internals through clear, stable contracts.

A module's public calling contract: inputs, outputs, errors, and synchronous or asynchronous behavior. Replacements must preserve it.

A memorable Internet name, usually registered for a renewable period. DNS supplies its records, while hosting runs the website.

A provider handling domain registration, renewal, and transfers. DNS can be hosted separately; compare renewal and multi-year costs.

A server answering DNS queries. Setting authoritative nameservers at a registrar delegates the domain's DNS to those servers.

A typed DNS entry: A/AAAA identify addresses, CNAME refers to a hostname, MX identifies mail servers, and TXT can carry verification data.

A provider of DNS, website proxying, and developer services. DNS-only records resolve names; proxying routes supported website traffic through its network.

Distributed infrastructure that delivers content and can cache reusable resources. Personalized data and authenticated responses need appropriate cache rules.

HTTP protected by TLS for encrypted transport and server identity verification. With a proxy, check both client-to-proxy and proxy-to-origin connections.

A secure gateway applying DNS, network, and HTTP policies to connected devices' traffic, useful for Internet filtering after traffic is connected.

An outbound connection established by cloudflared that lets Cloudflare reach a service on your infrastructure. The application and connector must stay running.

An application access service enforcing policies such as user identity. It can accompany Tunnel and needs explicitly configured access rules.

Cloudflare's application platform for static assets and server-side logic, with framework compatibility and runtime allowances to check.

A runtime model where the platform manages underlying server resources. Developers still own code, data, access, runtime constraints, and usage costs.

Storage that saves and retrieves file-like data by object identifiers, useful for images, attachments, and backups. Cloudflare R2 is one example.

A platform for website builds, deployment, and previews, including server-side capabilities. Check runtime constraints and plan eligibility.

A deployment environment for reviewing changes before production. Data and secret isolation still depend on its configuration.

Named configuration supplied through an environment, such as an API address. Values shipped to browser code are public; a variable name cannot protect secrets.

A platform offering application runtimes such as Machines for APIs and background processes, with compute, storage, and network resource charges.

An image packages a program and dependencies; a container runs it. Platforms can also start VMs from images. Persistent data requires separate planning.

Storage for data that must survive deployments. Fly Volumes have region and host constraints, do not automatically replicate between volumes, and do not replace backups.

A response-header mechanism telling browsers which origins may read responses, often needed for separate frontend and API domains. It does not replace API authentication.

Strictly the Linux kernel; everyday usage often means a complete distribution built around it with tools and repositories.

The core system component managing resources such as processes, memory and devices. It is not the graphical desktop.

Software that helps an operating system work with hardware. Check support for essential devices before changing systems.

An installable system combining a kernel, tools, installer and repositories, with its own versions, defaults and support lifecycle.

A tool for installing, updating and removing packages and handling dependencies. Tools and repositories differ across distributions.

A virtual computer running on a host, with a guest system on a dedicated virtual disk. Shared folders and device passthrough change its isolation boundary.

An image of installation media. Match architecture and verify origin and integrity; copying it onto an ordinary USB usually does not make that USB bootable.

A tool for multi-image boot media. After initializing a USB, copy supported ISOs and select them at boot. Initial installation formats the target media.

A firmware interface used before the OS loads, including initialization and boot selection. Its Secure Boot feature concerns trust checks for boot components.

A logical disk region that may hold system, user or boot data. Whole-disk erasure affects all partitions on the selected disk.

Structures and rules that organize files, directories and related information. Formatting creates a file system and can destroy existing data.

A processor that executes program instructions. Speed depends on cores, clock rate and workload, not a single number.

Working space for running programs, usually lost without power. It is not the disk that keeps files.

Longer-term storage for files. A file on disk does not imply that unsaved changes in a program have been written to it.

The process of initializing hardware, finding a boot target and loading the operating system. An app failing to open is not necessarily a boot failure.

System software that manages processes, files and hardware resources, such as Windows, macOS or Linux.

A running instance of a program. One application can use several processes; closing a window may leave background work running.

Several tasks making progress over a period of time, possibly by taking turns rather than running simultaneously.

Instructions that process inputs into results using explicit rules. AI-generated code still needs to be run and checked.

A way to express program rules. Choose based on the environment, ecosystem and existing project, not syntax alone.

A name used to refer to data in a program; its value may change as actions occur.

A test that decides which behavior to execute, such as whether a price is below a budget.

Repeating steps over items or while a condition holds, with a defined stopping condition.

Reusable processing that can accept inputs, return a result and sometimes produce side effects.

Information a program holds at a particular moment, such as whether a task is complete; changes need to be reflected in the interface.

An action or change, such as a button click. A program can react to an event by changing state; the event and state are distinct.

An interface for interacting with tools through text commands. Results also depend on the working directory and runtime environment.

The tools and environment needed to execute code. Versions and missing dependencies can change behavior.

An external library or tool a project needs. Track versions so the environment can be reproduced.

Version-control software for recording changes, comparing differences and recovering versions. It does not automatically verify correctness.

An identifiable state of code or a tool. Saving a version and testing its behavior complement rather than replace each other.

A unit of data sent over a network. Protocols specify how devices organize, send and handle it.

An address identifying a network interface on an IP network. It does not by itself identify which service should handle a request.

A number distinguishing network services on a device. After finding the IP, a client still needs the right listening port.

Wireless access to a local network. Being connected to Wi-Fi does not guarantee access to the internet.

A device connecting networks and choosing the next hop for traffic. Home routers often also provide Wi-Fi and address translation.

A system mapping names such as domains to network records. Record changes can be affected by caches and propagation time.

An application protocol for requests and responses between clients and servers; HTTPS adds encryption and authentication in transit.

The interface and interaction logic a user sees and operates. Browser code must not hold secret API keys.

Server-side software handling business logic, permissions or data. An API is an interaction contract, not the backend itself.

A number in an HTTP response describing the outcome category, such as success, not found or server error; details require the response and logs.

Placing a runnable program and its configuration in a target environment so intended users can access it.

Computing resources or software maintained by a provider and used over a network. Hosting does not remove responsibility for cost, access or data security.

A text format for structured data using objects, arrays and basic values. Valid syntax does not guarantee valid business data.

Storage such as localStorage lets the same browser and origin restore saved data after a reload. It is not account sync or a permanent backup.

A system for managing, querying and updating data, with deliberate choices about structure, access and backups.

Create, read, update and delete: four common data operations for translating interface requirements into data behavior.

A language for relational database queries and data operations. Different databases do not necessarily support identical SQL features.

An auxiliary data structure maintained for queries. It can speed up lookups at the cost of space and maintenance.

Multiple machines cooperating on data or tasks, with communication, failures and consistency to handle.

Processing work simultaneously. Splitting, transferring and combining work has overhead, so more machines are not always faster.

Stored reusable results that avoid repeated work. Define expiration or updates to prevent stale reads.

Using events and metrics to answer questions; data quality, measurement definitions and privacy boundaries affect conclusions.

The ability to adapt as users, data or work increase. Measure the bottleneck before choosing how to scale.

Adjusting model parameters from examples to make predictions or perform tasks; results depend on data and evaluation.

A model structure made of adjustable layers of computation, not a neuron-by-neuron copy of a human brain.

A model that learns patterns in data such as text and generates tokens in context. Fluent output does not guarantee factual accuracy.

Running a trained model on inputs to produce outputs, distinct from training and not a guarantee of factual verification.

Information available for a model's current task. Capacity is limited, so important requirements should be clear, relevant and current.

Plausible-looking output that is false or unsupported and needs checking against sources and actual behavior.

A system that can use tools across multiple steps toward a task. Define permissions, confirmation points and stopping conditions rather than assuming unlimited autonomy.

Comparing outcomes on tasks representative of actual needs with explicit criteria, not just model names, marketing or one demonstration.

Additional processing for complex tasks can increase latency and cost without guaranteeing correctness.

Maintained project facts, constraints and acceptance criteria that help AI understand the current work without treating old guesses as new facts.

Task instructions and context for AI, covering goals, current state, boundaries and expected results—not a formula that guarantees success.

Observable conditions for deciding whether a requirement is complete, including normal behavior, errors and existing behavior that must remain intact.

Steps covering triggers, inputs, processing, outputs and failure handling, best validated on a small scale before automatic execution.

The browser's object representation of a page's structure, which scripts can inspect and modify—not the pixels in a screenshot.

An interaction interface defining inputs, outputs and calling conventions. Use it with its authorization, limits and error-handling contract.

Turning raw content into structured data, handling and validating layout changes, missing fields and malformed input.

Controlling a browser by locating elements, waiting for page states and performing actions rather than guessing readiness with fixed delays.

Repeating an operation does not create additional unintended effects, especially important before retrying sends, writes or charges.