1. Upload multiple files at once
Drag 5 PDFs into one message. Ask a question that spans all of them. Claude holds the full context.
Use it for: comparing contracts, synthesising research papers, analysing a month of reports.
2. Ask Claude to ask you questions
Prompt: "Before you write anything, ask me the 5 questions that would make your output significantly better."
This surfaces your own assumptions and gives Claude the right context before burning tokens on a bad draft.
3. Style lock with an example
Paste your best past piece of writing and say: "Match this style exactly for everything I ask in this project." Claude extracts and mirrors tone, sentence length, vocabulary — far more reliably than adjectives alone.
4. Iterative refinement loops
After any draft: "Rate this output out of 10, explain the 2 biggest weaknesses, then produce a revised version addressing them."
Claude critiques and rewrites in one step. You get 80% of a second-pass review instantly.
5. System-prompt equivalent in Projects
Custom instructions behave like a system prompt. Put constraints here that apply to every conversation, not just the current one. Role, tone, output format, banned words — all of it.
6. Persistent memory via Projects
Claude doesn't remember across chats by default — but Projects share the same instruction set and file context. Use a project as a long-term memory layer by uploading an updated brief each week.
7. Code execution (Claude.ai Pro)
In the web UI, you can ask Claude to write and run Python code — and see the output. Upload a CSV and ask for a chart. No environment setup. No local Python install needed.