Tokens are the small units of text that AI processes, and context windows determine how much of your document an AI can consider at once—both factors directly impact what you pay per AI query. Understanding these constraints helps you design workflows efficiently: breaking large documents into chunks, requesting summaries instead of full processing, or choosing models with larger windows when you need to analyze long documents without fragmenting them.
Think of tokens like the memory capacity in a conversation. Imagine telling a friend your financial situation, and they can remember only the last 10 minutes of what you said. That's roughly how AI tokens work. Words get converted to tokens, and each model has a limit for how many it can hold at once.
This matters for single parents having long budget planning conversations. You might share your full financial picture early on, then ask detailed questions later—and the AI forgets the early details because it ran out of token space.
Tokens are small chunks of text. A token is roughly 4 characters, but it's not exact. The word "budget" might be one token. The phrase "single parent" is two tokens. Longer conversations use more tokens.
Every AI model has a context window—a limit to how many tokens it can remember in one conversation. ChatGPT's context window is roughly 4,000-8,000 tokens. Claude's is much larger, up to 200,000. This is a real difference in capability.
Imagine starting a conversation: "I make $3,200 monthly, my rent is $950, childcare is $850, and I want to save for my daughter's college." The AI remembers this. But if you have a very long conversation—asking 20 follow-up questions over hours—the AI might forget your original numbers and give contradictory advice later.
This isn't the AI being dumb. It's a hard limit on how much it can hold in active memory while generating responses. It's like a person who can hold a detailed plan in their head for 10 minutes but needs to write it down if they're planning for hours.
For long planning sessions, summarize at the start: "Here's my financial situation: [copy/paste all key numbers]. Keep these numbers in mind for this conversation." This puts your key data at the beginning of the conversation, where the AI is more likely to reference it.
For very long conversations about complex situations, break it into multiple sessions. Start fresh each time with a clear summary of your situation and what you're working on this session.
Or use Claude for long conversations, since its larger context window handles more information. If you're comparing multiple scenarios ("What if my income increased?" "What if childcare costs rose?"), Claude's size is an advantage.
Try this: Have a budget planning conversation with ChatGPT. Ask your question, get an answer, then ask 15 follow-up questions. Near the end, ask the AI to repeat your original income figure. Often it will be wrong or absent. This teaches you the token limit in action.
Peri can explain this concept, give practical examples, help you decide whether it applies to your situation, or recommend a journey if appropriate.
Explore related journeys or tell Peri what you're working through.