Lesson 04 · Claude Mastery Pro ~11 min read 3 workflows

Long-context: 200 pages, one prompt.

Claude can hold 200,000+ tokens in context — roughly 500 pages of text — and reason across all of it simultaneously. Most users never use this. They paste short snippets when they could paste entire books. This lesson covers the three workflows where long context genuinely changes what's possible.

The mental model

Long context isn't a bigger chat. It's a different kind of analysis.

When Claude can see your entire 80-page contract, your full quarterly report, or your complete codebase at once, the questions you can ask change fundamentally. Stop summarizing for it; let it see the whole thing.

Workflow 01 Multi-document synthesis

1

Compare 5 documents at once

Upload multiple documents in one conversation. Ask comparative questions.

The prompt that works

Synthesis promptsHere are 5 quarterly reports from competitors. Compare their AI strategies and tell me where each is investing. Here are 8 RFPs we've responded to in the past year. What patterns predict the wins vs. losses?

Best use cases

  • Competitive analysis (multiple competitor docs)
  • Pattern-finding across past customer feedback
  • Reviewing many contracts for non-standard clauses
  • Synthesizing research from many papers
Claude's reasoning quality degrades slightly at the extreme upper end of context. For 500-page comparisons, expect more general patterns; for 50-page comparisons, you get specifics.
Time savings: Hours of cross-document reading → 5-minute synthesis.

Workflow 02 Full-document interrogation

2

Paste the whole thing, ask anything

Upload a long document. Ask multiple specific questions across it without re-pasting.

The prompt that works

Interrogation prompts(Upload 80-page contract) Find every clause that limits our liability. Quote them with section numbers. What's the most unusual clause in this contract compared to standard SaaS agreements? If the customer doesn't pay, what are our remedies and timelines?

Best use cases

  • Contract review for non-standard terms
  • Earnings reports for specific metrics or commentary
  • Legal documents for specific clauses
  • Research papers for specific claims or methodology
Always verify Claude's citations against the actual source — long-context can occasionally hallucinate section numbers or quotes.
Time savings: 30 min of skim-reading → 2 min of targeted Q&A.

Workflow 03 Code understanding at codebase scale

3

Load a small codebase, understand it cold

For smaller codebases (under ~100 files), paste the lot into Claude and ask architectural questions.

The prompt that works

Code promptsHere's our entire backend. Walk me through how authentication flows from request to response. Where are the security-sensitive endpoints in this codebase? If I wanted to add a new payment provider, what files would I need to touch?

Best use cases

  • Onboarding to a new codebase
  • Code archaeology (understanding old code)
  • Architecture documentation generation
  • Pre-refactor analysis
Big codebases (1000+ files) exceed even Claude's window. Use Claude Code or codebase-aware tools for those.
Time savings: Hours of codebase spelunking → 10-min understanding.

Final challenge: pick the biggest document you have

Find a document you've been meaning to actually read carefully — a contract, a long report, a research paper, a manual. Paste it into Claude. Ask 10 specific questions that would have taken you an hour to find manually. Track time saved.

What you can do now

  • Upload entire documents (100+ pages) and ask questions across them
  • Synthesize patterns across multiple uploaded documents in one session
  • Use long-context for contract, research, or code analysis
  • Verify cited quotes/sections against the source for accuracy
Pro
Up next in Claude Mastery

Lesson 5 · Computer use: have Claude do tedious browser tasks

Setting up Claude's computer-use feature, the workflows worth automating (and the ones that aren't yet), safety and oversight patterns. See pricing →