Six ideas worth knowing

Tokens

A language model does not read text the way you do. It first breaks text into smaller pieces called tokens, which can be whole words, parts of words, or even punctuation. Everything the model reads and writes is measured in these pieces.

Context

A model answers using whatever information is in front of it right now: your instructions, the conversation so far, and any documents you have shared. Information you never provided is not available to it, no matter how obvious that information seems to you.

Generation

A generative model writes its answer one piece at a time by predicting what is likely to come next. It is not looking up a single stored answer in a database. That is why the same question can produce somewhat different answers on different days.

Grounding

Grounding means giving the model specific sources to work from, such as a document you upload or material a system retrieves for you. The model is then asked to base its answer on that evidence rather than on general recollection.

Agents

An agent combines a language model with instructions, tools, and several steps in a row. Instead of producing one reply, it can read files, run commands, check its own work, and keep going until a task is finished.

Verification

Confident, polished writing is not evidence that the content is correct. A model can produce a fluent answer that is wrong in small or important ways. Claims that matter should be checked against a source you trust.

Explore a concept

Press the button to see one of the six concepts, along with a question you could try asking an AI assistant.

Your concept will appear here.

Ideas for your fork

This site is deliberately unfinished. Here are some things you could ask Claude Code or Codex to do: