Question 1

What is a token and a context window?

Accepted Answer

A **token** is the unit a language model processes — roughly a word fragment — and the **context window** is how many tokens it can consider at once. Both determine cost, capability and the limits you run into.

**What a token is.** Not a word and not a character. Text is broken into pieces by a **tokeniser**: common words are usually one token, rarer words split into several, and punctuation and spaces are tokens too. A frequently cited approximation for English is around **three-quarters of a word per token**, or four characters.

**Why tokenisation is not neutral:**

**Languages differ substantially.** Tokenisers trained predominantly on English split other languages into more tokens for the same meaning — so the same text costs more and consumes more context in some languages than others, which is a genuine inequity.

**Code, numbers and unusual strings** tokenise inefficiently.

**It explains specific failures.** Character-level tasks — counting letters, reversing words, handling spelling — are hard because the model does not see characters. This is why questions about letter counts produce surprising errors from systems that handle complex reasoning well.

**The context window.** The maximum tokens the model can attend to in one request, **including both input and output**. Everything the model can use must fit: system instructions, conversation history, supplied documents, and the response being generated.

**What happens at the limit.** Older material is truncated or summarised. This is why a long conversation appears to forget earlier detail — it is no longer present, rather than being forgotten.

**Why bigger windows are not simply better:**

**Cost and latency** rise with length.

**Attention degrades across long contexts.** Research has repeatedly found models perform worse on information in the **middle** of a long context than at the beginning or end — sometimes called "lost in the middle" — so filling a large window is not equivalent to the model using it all well.

**More context can dilute** the relevant material.

**The practical implication:** retrieving the relevant portion and supplying that is usually better than supplying everything, even where everything fits.

Question 2

What is a large language model and how is it different from a search engine?

Accepted Answer

A large language model (LLM) is a statistical model trained to predict likely continuations of text. A search engine retrieves documents that exist. They answer questions in fundamentally different

Question 3

How does speech recognition work?

Accepted Answer

By converting sound into a numerical representation, then using a model trained on enormous quantities of speech to produce the most probable sequence of words — and "most probable" is the key,

Question 4

What is an API key, and how is it different from an OAuth token?

Accepted Answer

An API key identifies an application; an OAuth token represents a user's authorisation for an application to act on their behalf. They answer different questions, and using one where the other is

Question 5

Why do AI models make things up?

Accepted Answer

Because a language model is trained to produce plausible continuations of text, not to state truths — and fluent plausible text and accurate text are different targets that usually coincide and

Question 6

What does digital accessibility actually require?

Accepted Answer

That people with disabilities can perceive, operate, understand and use a digital service — and the practical standard almost everywhere is WCAG, whose four principles are exactly those four words.

Question 7

How to choose a programming language for beginners?

Accepted Answer

Choosing a programming language as a beginner involves identifying goals, considering community support, evaluating syntax, and exploring job opportunities. Experimenting with multiple languages can also aid in making the right choice.

Related Questions