Context Window
Everything the AI can 'see' at once: your prompt, the files, the images, and the conversation so far. When it fills up, older details fall out of view.
What it is
The context window is everything the model can hold in view at one moment: your current prompt, the files it has read, the images you pasted, and the conversation up to now. It has a fixed size. Once it is full, the model can no longer see all of it, and the oldest or least relevant parts effectively drop away.
Simple analogy
It is the model’s desk, not its memory. Only what fits on the desk right now can be used. Pile on too much and things slide off the edge.
Why this matters for designers
Almost every “the AI forgot what we agreed” or “it stopped following my rules” moment is a context-window problem. If you paste a giant design system and then have a long conversation, the early instructions can fall out of view. The fixes are practical: keep durable rules in a CLAUDE.md file the model re-reads, clear context between unrelated tasks, and avoid dumping more than the task needs.
How it works in practice
- Everything you send counts toward the window, including images and prior messages.
- The bigger the pile, the more it costs and the easier it is to lose detail.
- Smaller, focused context produces sharper, cheaper, more reliable output.