A pharmaceutical company sends a 150K-token clinical trial report to Claude for analysis. The report has: the trial conclusion (most critical) in section 5 (middle), patient demographics in section 1 (beginning), and statistical methodology in section 8 (end). Claude correctly summarizes the demographics and methodology but provides an incomplete analysis of the trial conclusion. The team realizes the conclusion's position is problematic. Where should the most critical information be placed in a long document sent to Claude?
Document Structuring
Loading lesson content...
A fintech analytics team sends a 200K-token financial report containing 40 pages of pivot tables, balance sheets, and quarterly comparisons. The model misreads several data points, for example, it reports Q3 revenue as 4.2M when the table clearly shows 2.4M. The data is structured as: 'Q1 3.1M Q2 3.8M Q3 2.4M Q4 4.2M' in plain text. How should the team format their tabular data for best model comprehension?
A developer tooling team is building an AI code review assistant that analyzes pull requests. Each PR includes 15-30 changed files across multiple directories. Currently, the team concatenates all file contents into a single text block without file markers. The model frequently attributes code to the wrong file, for example, it says 'the error handling in src/utils/auth.ts needs improvement' when that code is actually in src/middleware/validation.ts. What format should the team use for presenting code files in context?