Context Window Optimization for ADHD Working Memory Constraints

Context window refers to the amount of information an AI model can hold and reference simultaneously—essentially its working memory. For neurotypical users, this is a technical parameter. For neurodivergent learners, especially those with ADHD, it's a critical accessibility feature that directly impacts how effectively you can use AI as an external cognitive scaffold.

Your brain's working memory has finite capacity. ADHD often reduces this further, making it harder to juggle multiple pieces of information at once. When you dump everything into an AI conversation, you're asking the model to do what your brain struggles with—hold it all in active working memory. But here's the key insight: AI context windows aren't like your brain. They don't fatigue. You can strategically structure conversations to match your actual cognitive capacity, not your aspirational capacity.

How Context Windows Work in Practice

Most modern LLMs (large language models like Claude or GPT-4) can handle 100,000+ tokens of context—roughly equivalent to 75,000 words. That sounds limitless, but it's not about raw capacity. It's about retrieval efficiency. The more context an AI has, the more it must search through to find relevant information. Early tokens (information you provided first) actually become slightly less influential in decision-making, a phenomenon called "lost in the middle."

For ADHD users, this creates a practical problem: if you dump your entire project, all your notes, and all your constraints into one conversation, the AI model may lose track of your core focus. Worse, you've created cognitive overload for yourself—you have to manage and reference everything you've shared.

Strategic Segmentation Approach

Instead of monolithic conversations, segment your work into context-bounded sessions. Dedicate one conversation thread to breaking down a project, another to drafting, another to editing. This mimics how people with lower cognitive load naturally work—one task at a time—but with AI as your external memory.

Within each conversation, establish a "system context"—the first message that sets constraints. Include only what the AI must know: your specific learning difference ("I have ADHD and need information chunked into 3-sentence maximum explanations"), the exact task, and success criteria. Reference external documents (like a brief project overview in Notion) rather than pasting them.

This isn't about limiting the AI's knowledge—it's about creating signal-to-noise optimization. You're teaching the AI to filter information the way your brain needs it filtered.

Practical Implementation Trade-offs

The trade-off: segmented conversations require you to manage multiple threads and carry context between them. Some people with executive function challenges find this harder than one monolithic conversation. Test both approaches. If you choose the single-conversation model, at minimum use clear section breaks ("---") and numbered topics to help both you and the AI maintain focus.

When managing chronic hyperfocus sessions (common in ADHD), be aggressive with context management. Every 2-3 hours, start a fresh conversation with a summary prompt: "Here's what we've completed: [bullet points]. Next focus: [one specific task]." This prevents context degradation and forces you to consolidate learning.

Try this: Take your next complex project and deliberately create three separate ChatGPT or Claude conversations: one for planning/breakdown, one for execution, one for review. In each conversation's first message, write one sentence about your learning difference and exactly what you need from that phase. Notice whether breaking work into context-bounded sessions feels like helpful structure or additional friction—that answer tells you your optimal AI workflow architecture.