A dialogue, not a one-off question

An AI assistant remembers what you discussed within the current chat. This is called context. Thanks to it you can refine and build on an idea without repeating everything — like talking to someone who listened from the start.

In the second message you never repeated the word "borscht" — the model inferred it from context. That's the power of dialogue: short refinements instead of long repeats.

How to think about context

Picture context as a "desk": everything you put on it in this chat, the model sees and uses. You can accumulate details step by step:

The model remembers the topic (exercise), the audience (beginners) and the tone (friendly) — you don't repeat it every message.

Key nuances

• A new chat is a blank slate. Starting a new dialogue drops the previous context. Handy when the topic changed and old details only get in the way.
• Context isn't infinite. In very long chats early details may "fall out" of attention. If something matters, restate the key point closer to your current message.
• Mixing topics hurts. Discussing recipes and a work report in one chat can make the model confuse contexts. Keep different tasks in different chats.
• Memory across chats isn't available everywhere and is usually configured separately. Don't rely on it by default.

A common mistake

Someone gets a good answer, then 20 messages later on another topic asks "polish that text" — and is surprised the model "forgot". In reality the fragment simply drowned in a long context. Fix: paste the text again or start a fresh chat with it.

Exercise

Start a dialogue: ask for a gift idea for a friend. Then, without repeating the intro, refine one message at a time: "he loves cooking", "budget under $40", "need it by Friday". Watch the model hold all conditions at once.

Rule: one dialogue, one task. That keeps context clean and answers precise.