How do I ask an LLM to transcribe handwritten financial notes?

I have pages of handwritten financial notes (images). How do I get an LLM to transcribe them cleanly into org-mode?

Start a fresh chat (no memory) and paste the following prompt, then upload the note images one page at a time.

You will receive a series of handwritten note pages (images). Your
task is to:

1. This is a new chat. Do not use your memory.

2. Perform accurate OCR transcription.

   - Clean up handwriting.
   - Preserve meaning precisely.
   - Expand abbreviations only when obvious (e.g., "CVA" -> "Credit
     Valuation Adjustment").

3. Convert the content into a clean, simple Emacs org-mode document,
   using:

   - * for main sections. Use normal capitalisation for
     titles/headers.
   - ** for subsections. Use normal capitalisation for
     titles/headers.
   - - bullet lists
   - no LaTeX unless handwritten
   - no extra commentary

4. If the page contains something resembling a diagram, box, or
   flowchart, represent it as a separate PNG image. The diagram
   content should be clearly structured, and the background must be
   white.

5. Do NOT summarise. Transfer the meaning exactly as written, cleaning
   only grammar and structure where absolutely necessary.

6. Each page should be converted independently, as if it were a
   section of one bigger .org file.

7. The context of the notes is Financial Markets. Keep terminology
   intact (e.g., "wrong-way risk", "funding rolls", "novation", etc.).

8. Output formatting - strict:

   - ONLY output the raw org-mode text.
   - DO NOT wrap the output in Markdown code blocks.
   - DO NOT include any conversational text, introductions, or
     conclusions.
   - Start directly with the first header or content line.

Do not output any content which is not provided from the image to OCR.

Do you understand these instructions?

Manual usage across multiple LLM chat interfaces.

The output is draft content — generic org-mode, not current format. If the notes are destined for the repo, a human must reshape the output to satisfy the frontmatter contract before committing. See How do I create a new doc? for the codegen scaffold, and Document types for the per-type section requirements.

• LLM prompt recipes — index of all LLM prompts.
• How do I create a new doc? — scaffold the target doc after transcription.
• Document types — frontmatter and section contract to apply.