Why this matters
The biggest leap most people make in AI literacy is when they stop typing every question fresh and start using persistent context — projects, custom instructions, file uploads. This module covers how to do that across the major tools.
Projects and custom instructions
A project (or workspace, or “Gem”) is a saved environment with persistent instructions and uploaded files. It means you don’t have to re-explain who you are or what you’re working on every time.
Across tools:
- ChatGPT Projects — Save instructions and files. New chats inside the project inherit both.
- Claude Projects — Same idea. Excellent for long-running work.
- Gemini Gems — Custom assistants you can configure and reuse.
Custom instructions are persistent rules that apply to every conversation. Most tools let you set:
- Who you are (role, expertise, context)
- How you want responses formatted (length, tone, structure)
- What to avoid (banned phrases, off-topic content)
Setting good custom instructions once saves hours over time.
Uploading documents
Modern AI tools accept:
- PDFs — research papers, reports, contracts, books
- Spreadsheets (CSV, XLSX) — data analysis, summary, transformation
- Images — diagrams, screenshots, photos for analysis
- Word docs, text files, slides — most formats work
Once uploaded, the model can read, reference, and analyze the content like any other context.
One thing worth knowing: when you upload an image or a scanned PDF, the model uses OCR (Optical Character Recognition) to “see” and read the text inside the image. So a photo of a receipt, a screenshot of a slide, or a scan of an old document all become readable content the model can work with.
Summarization, extraction, and analysis workflows
Common file workflows:
Summarization
- “Summarize this document in 3 bullet points.”
- “Give me an executive summary in 100 words.”
- “What are the main arguments and counterarguments?”
Extraction
- “List every action item in this document.”
- “Extract all names, companies, and dates.”
- “Pull out every statistic with its source.”
Analysis
- “What assumptions does this document rely on?”
- “Identify gaps in this argument.”
- “Compare this to [other uploaded document].”
Transformation
- “Convert this report into a one-page slide outline.”
- “Rewrite this policy document as a checklist.”
- “Turn this into a FAQ.”
When to upload vs when to paste
Upload when:
- The document is long (over a page or two)
- It’s already formatted (PDF, spreadsheet, presentation)
- You’ll reference it across multiple conversations
- You want the model to see images, charts, or layout
Paste when:
- The content is short (a paragraph, a few bullet points)
- You want to control exactly what’s included
- You’re combining text from multiple sources
- The source is a webpage and you want to strip the noise
A good rule: if you’d put it in a project, upload it. If you’re using it once, paste it.
Key takeaways
- Projects + custom instructions = persistent context, no repetition
- Upload files for anything long, formatted, or reused
- Paste for quick one-off information
- Summarize, extract, analyze, transform — file workflows cover most knowledge work
- Set custom instructions once, save hours forever
Quick Check
1. A "Project" (or Workspace, or Gem) is best described as:
2. Custom instructions are useful because they:
3. OCR (Optical Character Recognition) is what lets a model:
4. You should upload a file (rather than paste content) when:
5. Which is NOT a typical file workflow?