Glossary
The AI, software, and web terms we use with clients, defined in plain language and tied to how we actually ship.
AI systems
- AI agentAn AI agent is an AI system, usually built on an LLM, that takes a goal, plans the steps to reach it, calls tools or APIs along the way, and acts with some autonomy. Unlike a chatbot that answers one prompt, an agent runs a loop until the task is done.
- EmbeddingAn embedding is a list of numbers that represents the meaning of a piece of text, image, or other data, so that similar things end up close together in that numeric space. Embeddings are what let a vector database find content by meaning rather than by exact words.
- Fine-tuningFine-tuning continues training an existing language model on your own labelled examples so it adopts a specific tone, format, or task more reliably. It changes the model's weights, unlike RAG, which leaves the model untouched and supplies knowledge at query time.
- HallucinationAn AI hallucination is when an LLM states something false but plausible-sounding, and does it with confidence. The model is predicting likely text, not checking facts, so it can invent a citation, a number, or a feature that does not exist while sounding completely sure of itself.
- LLMA large language model (LLM) is an AI model trained on vast amounts of text to predict the next token, which lets it generate and transform language: writing, summarising, classifying, answering questions, and calling tools. It is the engine behind most modern AI features.
- Prompt engineeringPrompt engineering is the practice of writing and structuring the instructions and context you give an LLM so it returns reliable, useful output. It covers how you phrase the task, what examples and data you include, and how you format the request, all tuned through testing rather than guesswork.
- RAGRetrieval-augmented generation (RAG) is a technique where a language model fetches relevant passages from your own data at query time and uses them to ground its answer, so responses cite real, current sources instead of relying only on what the model memorised in training.
- Vector databaseA vector database stores text, images, or other data as numerical vectors (embeddings) and finds the most similar items by meaning rather than exact keywords. It is the retrieval engine behind most RAG systems, letting an AI feature pull the passages most relevant to a question.
Building software
- APIAn API is a contract that lets two software systems talk to each other in a defined way. It exposes a set of requests one program can make and the responses it gets back, so apps can share data and features without knowing how the other side is built internally.
- CachingCaching is storing the result of expensive work so the next request can reuse it instead of recomputing. Done well it makes software dramatically faster; done carelessly it serves stale or wrong data, which is why cache invalidation is famously hard.
- CI/CDCI/CD (continuous integration and continuous delivery) is the practice of automatically building, testing, and shipping code every time it changes, so problems surface in minutes and releases are routine rather than risky events.
- Code reviewA code review is a second engineer reading a change before it ships, to catch bugs, share context, and keep the codebase consistent. It is one of the cheapest ways to raise quality, because a fresh pair of eyes sees what the author cannot.
- Design systemA design system is a shared library of reusable UI components, tokens, and rules that keep a product visually consistent and faster to build. It is what stops every screen from reinventing buttons, spacing, and colour.
- Feature flagA feature flag is a switch in your code that turns a feature on or off without a new deploy. It lets you ship work hidden, release it to a few users first, and turn it off instantly if something goes wrong.
- Laravel OctaneLaravel Octane boots a Laravel application once and keeps it resident in memory using a high-performance server such as Swoole or FrankenPHP, serving requests without rebooting the framework each time. The result is typically several times more throughput and much lower latency.
- MonorepoA monorepo is a single repository that holds many projects, packages, or services together, instead of splitting them across separate repos. It makes sharing code and changing several things at once easier, at the cost of needing better tooling to stay fast.
- MVPA minimum viable product (MVP) is the smallest version of a product that delivers real value to early users and tests your riskiest assumption. The point is not to ship something cheap, it is to learn the truth about demand with the least time and code.
- ObservabilityObservability is how well you can understand what a running system is doing from the outside, through logs, metrics, and traces. Good observability means that when something breaks, you can find out why in minutes instead of guessing.
- RefactoringRefactoring means improving the internal structure of code without changing what it does from the outside. The product behaves the same for users, but the code becomes cleaner, clearer, and easier to extend. It is how teams pay down technical debt so the next feature is faster and safer to build.
- SaaSSaaS, or software as a service, is software delivered over the web as a subscription. The customer logs in through a browser and pays monthly or yearly, while the provider runs the servers, updates, and backups. There is nothing to install and no infrastructure to maintain on the customer side.
- Tech stackA tech stack is the set of programming languages, frameworks, and tools a product is built with, for example React for the screen, Node for the server, and Postgres for the database. The choices fit together as layers, and together they shape how the software is built, run, and changed over time.
- Test-driven developmentTest-driven development (TDD) is a practice where you write a failing test for a small piece of behaviour first, then write just enough code to pass it, then clean up. The tests come before the code, which keeps the design honest and the safety net thorough.
- Web appA web app is software you use inside a browser instead of installing it on your device. Unlike a plain website, which mostly shows you information, a web app lets you do work: log in, enter data, and get results back. Gmail, a dashboard, or an online editor are all web apps.
- WebhookA webhook is an automated message one system sends to another the moment an event happens, as an HTTP request to a URL you provide. Instead of repeatedly asking "anything new yet?", your app is notified the instant something changes, such as a payment, a new signup, or a status update.
Web & growth
- Core Web VitalsCore Web Vitals are a set of Google metrics that measure real-world page experience: mainly loading speed (LCP), responsiveness to input (INP), and visual stability (CLS). They put a number on how fast and steady a page feels to an actual visitor, not just on a test machine.
- SEOSEO, or search engine optimization, is the practice of making a site easier for search engines to find, understand, and rank, so the right people discover it without paid ads. It pulls together three things: useful content, technical health like speed and structured data, and links from other sites.
- Structured dataStructured data is machine-readable markup you add to a page, usually schema.org vocabulary written as JSON-LD, that tells search engines exactly what the content is: an article, a product, a business, a defined term. It turns a page they have to guess about into one they can read with certainty.
Working with a studio
- CTOA CTO (Chief Technology Officer) is the senior leader who owns a company's technology: the strategy, the architecture, and the engineering team. At an early startup this is often the technical co-founder. Not every company is ready for a full-time hire, which is why the fractional CTO exists.
- Fractional CTOA fractional CTO is an experienced Chief Technology Officer who works part-time or on demand, for one company or several, instead of being a full-time hire. You get senior technical leadership, the strategy and architecture and hiring calls, for a few days a month rather than a full executive salary.
- Nearshore developmentNearshore development means outsourcing software work to a nearby country in a similar timezone, rather than a distant one. Compared with far-off "offshore" work, the overlap in working hours and culture makes everyday collaboration far easier: same-day answers, real meetings, fewer things lost in translation.
- Product-market fitProduct-market fit (PMF) is the point where a product genuinely satisfies a strong market demand: people want it, use it, pay for it, and tell others. Before it, growth is a grind and nothing quite sticks. After it, the main problem becomes keeping up. Finding it, with the least time and code, is the whole reason an MVP exists.
- Scope creepScope creep is when a project's requirements keep growing past what was agreed, one small "while you're at it" at a time, until the timeline and budget blow up. Each addition sounds reasonable on its own; together they sink the deadline. Fixed scope and a signed scope document exist precisely to stop it.
- Technical debtTechnical debt is the future cost of shortcuts taken to ship fast today. Like financial debt it charges interest: changes get slower, bugs get more frequent, and onboarding new engineers takes longer. It is normal and often worth it, as long as you pay it down deliberately with refactoring instead of pretending it is free.
- Technical due diligenceTechnical due diligence is an expert audit of a company's codebase, architecture, team, and risks, usually run before an investment or acquisition. It answers the question money cares about: is this software actually worth what people think, or is it held together with hidden technical debt and security risk?