Sarah is a lawyer at a mid-size firm. Administrative law. Dense briefs, case files arriving in chunks, clients who want to take it all the way. One morning she receives a heavy case: 60 pages of administrative filings, previous resolutions, technical reports, and a cross-referenced email chain that can’t seem to agree with itself.
She opens a conversation with Claude and does what seems logical: pastes all the documentation upfront and starts working. She sets the tone — precise, measured, no ornament. She maps the legal strategy. She marks which arguments are central and which are secondary. For nearly an hour, the conversation flows. The responses are strong. The proposals make sense.
On message 26, something shifts. Claude starts using phrasing she’d explicitly ruled out at the beginning. It reframes a core argument with a different angle. In one reply, it seems to contradict a clear instruction Sarah gave early on. Her first instinct: the AI is becoming inconsistent. Something’s broken.
Nothing was broken. The conversation simply didn’t fit on the desk anymore.
What Is the Context Window and Why It Doesn’t Remember Everything
The context window in Claude is like a lawyer’s desk. It holds papers, folders, sticky notes, a laptop, maybe a coffee. As long as everything fits and stays visible, you can work with clarity. You know what’s important, what’s pending, what’s already been decided.
Now imagine you keep stacking documents without removing anything. There comes a point where the desk is the same size, but the material no longer fits. You start covering things up. The last thing you placed stays on top. The first thing you put down disappears under a pile.
Every conversation with Claude has a context window: a limited space that holds everything the model is considering when it responds to you. It’s not long-term memory. It’s working memory. While the conversation fits in the window, Claude remembers your instructions, tone, decisions, and relevant data. When it fills up, there’s no warning alarm. It just starts prioritizing what’s most recent.
It doesn’t «forget» all at once. It stops seeing what got buried.
Tokens Explained Without the Drama
Understanding the size of that desk requires a word that sounds more intimidating than it is: token. A token isn’t exactly a word and isn’t exactly a letter. It’s a unit of text. In practice:
- A page of written text typically runs about 500 tokens.
- An 80-page PDF can take up around 40,000 tokens.
You don’t need to count them or optimize them like calories. Just understand that they exist and that the space is finite. Each model has a different window: Opus can hold more context than Sonnet, and Sonnet more than Haiku. But none of them remembers «everything forever» within a single conversation.
When Sarah pasted 60 pages at the start and then kept a long conversation going, she didn’t do anything wrong. She just filled the desk very quickly.
Signs Your Conversation Is Saturated
When the context window fills up, Claude starts doing something very human: it pays more attention to whatever you said most recently. The consequences are predictable:
- Rules you set early on start to dilute.
- The tone may shift without you noticing.
- Small contradictions start appearing.
- Claude repeats ideas that were already settled.
There are reliable warning signs: Claude ignores a clear rule you gave earlier, returns a correct answer that no longer fits with previous decisions, or asks about something that was already resolved. When you see two or three of these signs together, don’t push through. It’s not a prompt engineering problem. It’s a space problem.
Three Fixes That Work Without Starting Over
1. Summarize and Reopen (Most Effective)
Ask Claude to produce a structured summary of the conversation: facts established, decisions made, rules still active. Take that summary and open a fresh conversation. It acts like a clean desk with only the essentials on it. You’re not starting over. You’re continuing with order.
2. Reposition Critical Instructions
If you can’t close the conversation yet, move what matters to the end. Important instructions work better when they’re close to the last question. Not because they’re more important inherently, but because they stay visible within the window. Don’t repeat everything. Reaffirm what’s essential.
3. Chunk Before You Work
Instead of pasting a massive PDF and working on top of it, change the order: first ask for a summary, then decide which parts are relevant, and only then dive deep. That way you use the window for thinking, not for storing unfiltered text.
Closing a Conversation Well Is Also Work
One of the most productive habits is knowing when to close. When a conversation has served its purpose, don’t stretch it «just in case.» Extract what matters, save the state, and open a new one when it’s time. Working with AI isn’t about maintaining an eternal thread. It’s about managing sessions. You won’t notice this on day one. You’ll notice it after ten long projects and none of them has turned into chaos.
Claude’s context window isn’t an annoying limitation — it’s a design feature that, when managed well, forces you to be more precise and organized in how you delegate work. And that, paradoxically, improves your results.
This is just a taste. The full book shows you how to turn AI into your most productive team member.
📖 Your Digital Employee
Claude and AI as your best collaborator