
**TL;DR** - Productive procrastination: why doing unimportant things feels good; Anthropic cache downgrade; how AI context windows affect what you can build.
1. If using Claude for long documents, evaluate cache hit rate and whether it worth the cost
2. Audit your own productive procrastination patterns - AI makes it easier to appear productive while avoiding real work
3. When designing AI products, treat context window as architectural constraint, not just a feature
**Example Prompt:**
Calculate cost difference between processing a 500-page document in a single call vs chunked processing with RAG.
| Pros | Cons |
|---|---|
| Cache changes well-documented by Anthropic | Cache optimization complex and error-prone |
| Context window thinking architecturally important | Chunked processing introduces its own quality issues |
|---|---|
| Productive procrastination worth naming and confronting | Awareness alone does not fix the behavior |