AI & LLM Basics: A Plain-English Guide

If you've been nodding along when people talk about "AI" while privately wondering what on earth is going on — this one's for you. No jargon, no overwhelm. Just a clear, honest explainer you can come back to whenever you need it.

What is AI?

AI stands for Artificial Intelligence — software designed to do things that usually require human thinking.

Here's something that might surprise you: AI isn't new. It's been around for decades. Your email spam filter is AI. Netflix's recommendation engine is AI. The facial recognition that unlocks your phone is AI. Even autocorrect (for better or worse) is a form of AI.

What's changed is the type of AI we're all talking about now — and how visible and accessible it's become.

What's Generative AI?

"Generative AI" (or Gen AI) is the term you'll hear constantly in the news — and it describes something genuinely different to older AI.

Traditional AI is mostly about recognising things: detecting spam, recommending content, identifying objects in photos. It works with what's already there and categorises it.

Generative AI creates things. It produces new content — text, images, audio, video, code — that didn't exist before. Instead of just spotting patterns, it can draft you a proposal, generate an image from a description, or think through a problem with you.

The moment it went mainstream? November 2022, when OpenAI launched ChatGPT. Within five days it had a million users. Within two months, 100 million. Nothing before it had made AI this visible — or this accessible — to everyday people.

The technology behind tools like ChatGPT and Claude is called a Large Language Model (LLM). LLMs are trained on enormous amounts of text — books, websites, articles, conversations — which is how they've learned to understand and respond in a way that feels surprisingly natural.

The short version: Generative AI is the current wave of AI that can make things, not just analyse them. LLMs are the text-based flavour powering most of the tools you've heard of.

What's Agentic AI?

If Generative AI is about creating things, Agentic AI is about doing things.

"Agentic" describes AI that can take a sequence of actions, make decisions along the way, and complete multi-step tasks — without needing you to prompt it at every single step.

Most AI tools you've used so far are conversational: you ask, it responds. One exchange at a time. Agentic AI is different — it can receive a goal and then work through a whole series of steps to achieve it, using tools, checking results, adjusting course if something doesn't go to plan, and looping back until the job's done.

What does that look like in practice?

• Searching the web, reading pages, and summarising findings — without you lifting a finger between steps

• Writing code and then running it to check it works, then fixing errors automatically

• Managing files, drafting documents, and coordinating tasks across multiple apps

• Completing workflows that would otherwise take you half an hour to step through manually

In short: Generative AI responds. Agentic AI acts.

This is the direction most AI tools are actively moving in — and it's already here. Claude Cowork is a direct example: it's agentic AI built for non-developers, designed to take action on your computer and work through real tasks, so you don't have to manage every detail yourself.

The main companies — who's who

OpenAI — the team behind ChatGPT. Currently the most well-known AI company globally.

Anthropic — the team behind Claude. Founded by researchers focused on building AI safely. (Anthropic's AI is what powers the tools we use at digiDen Studio.)

Google DeepMind — Google's AI research division. They make Gemini, which powers AI features across Google's products — including Search.

Meta AI — Facebook and Instagram's parent company. They build Llama, an open-source family of models that developers can build on, plus Muse Spark, their newer proprietary model.

Mistral — a French AI company making smaller, efficient models popular for business and developer use.

xAI — Elon Musk's AI company. They make Grok, available via X (formerly Twitter).

Most of these offer both a consumer chat tool (for everyday personal use) and an API — which lets businesses build AI into their own products and services.

OpenAI's current models — what's the difference?

OpenAI's models have moved well beyond the familiar "GPT-4" era. As of June 2026:

• GPT-5.5 Instant — the current default in ChatGPT. Fast, conversational, and significantly more accurate than earlier versions. What most users interact with day-to-day.

• GPT-5.4 mini — lighter and cheaper. Good for simpler tasks where speed matters more than depth.

• GPT-5.3-Codex — OpenAI's specialist coding model, combining code generation, reasoning, and general intelligence. Aimed at software developers.

The line-up evolves quickly and the naming can feel chaotic. That's not just you.

Claude's current models — what's the difference?

Claude (by Anthropic) uses a tiered model range — a sliding scale of speed vs. power:

• Haiku — fast and efficient. Built for simple tasks and quick answers.

• Sonnet — the all-rounder. Great balance of capability and speed; what most Claude users interact with by default.

• Opus — highly capable. Better for complex reasoning and work where quality matters most.

• Claude Fable 5 ✨ — Anthropic's newest and most powerful model, just released. It sits above Opus in a new "Mythos" tier and excels at software engineering, knowledge work, and complex vision tasks. Available to Claude Pro, Max, Team, and Enterprise users.

Worth knowing — access update: Shortly after Fable 5 launched, the US government issued an export control directive requiring Anthropic to suspend access to both Fable 5 and the related Mythos 5 model for all foreign nationals — which includes UK users. Anthropic disabled both models globally to comply. The stated reason is a reported security vulnerability that the government believes could be exploited; Anthropic reviewed it and said the actual risk was minor and already known in other models. Either way, the directive stands for now. The good news: this applies only to these two top-tier models. Haiku, Sonnet, and Opus are completely unaffected — so if you use Claude for everyday writing, research, or planning, nothing changes. It’s an unusual moment: the first time the US government has used export controls to restrict access to a commercially available AI tool. Worth keeping an eye on as it develops.

A Dummy's Guide to Claude products

Anthropic makes a few different products that are easy to mix up:

Claude AI / Claude.ai
The chat interface — available on web, mobile, and desktop. You type, Claude responds. This is the "front door" to Claude and what most people mean when they say "I use Claude". Great for writing, brainstorming, research, summarising, and planning. Built for everyone.

Claude Cowork
A desktop tool for non-developers. Claude Cowork is agentic AI in action — it lets Claude take real action on your computer, managing files, working across your apps, and coordinating tasks. Built for people who want AI to actually do things, not just answer questions. Think of it like a capable assistant who can open your apps, read your documents, and work through tasks alongside you — no coding knowledge needed.

Claude Code
A command-line tool for developers. Lets software engineers delegate coding tasks — writing, reviewing, and debugging code — directly from their terminal. Designed for people already comfortable working in code.

TL;DR: Claude.ai is for everyone. Cowork is for non-technical people who want AI to take action. Code is for developers.

Quick reference: Glossary 📖

Plain-English definitions for the terms that keep coming up.

AI — Artificial Intelligence
Software designed to perform tasks that usually require human intelligence — reading, writing, pattern recognition, decision-making. Has existed for decades; the current wave is just much more visible.

LLM — Large Language Model
The technology behind conversational AI tools like ChatGPT and Claude. Trained on vast amounts of text to understand and generate language naturally.

Generative AI
AI that creates new content — text, images, audio, video, code. The current wave of AI tools most people are using. ChatGPT, Claude, and image generators like Midjourney are all examples.

Agentic AI
AI that can take sequences of actions, make decisions, and complete multi-step tasks on its own — rather than waiting for a prompt at every step. The key difference: Generative AI responds to you; Agentic AI acts for you.

SEO — Search Engine Optimisation
Making your website content rank well in traditional search results (Google, Bing, etc.). Focuses on keywords, structure, and links.

GEO — Generative Engine Optimisation
Optimising your content so it appears in AI-generated summaries — like Google's AI Overviews or Perplexity's answers. A newer discipline that builds on SEO.

AEO — Answer Engine Optimisation
Optimising your content to be the specific answer that AI assistants give when someone asks a question. Closely related to GEO — the two terms are often used interchangeably.