How to Talk to AI Citizens: Context Is Everything

Get complete working code in minutes instead of spending 6 hours debugging

3-5xFaster Completion
8+Hours/Day with AI
10+Active Projects

🎯Why This Matters

You'll spend 8+ hours every day asking AI citizens (Rafael, Sofia, Emma) for code generation, bug fixes, quality verification, and proposal help. The difference between effective and ineffective questions is dramatic:

❌ Bad Question
Vague request → AI guesses wrong → 6 hours wasted debugging
✅ Good Question
Full context → Exact solution → 2 minutes to working code

⭐The Golden Rule

AI Citizens Work Across 10+ Projects
Rafael, Sofia, and Emma don't have your context. They need to know: Which project? What goal? What error? What did you try? Without this, they're guessing.

📋How to Give Context (The Right Way)

Don't explain what you think the AI needs to know. Show them the actual source material:

🥇 Best
Copy-Paste Content Directly
"Here's the error I'm getting:

TypeError: Cannot read property 'map' of undefined
at UserList.tsx:42:18
at renderWithHooks
Here's the component code:

const UserList = ({ users }) => {
  return users.map(user => ...
}"
Why: AI sees exactly what you see. No ambiguity. Instant understanding.
🥈 Good
Give File Pointers
"I'm getting an error in /src/components/UserList.tsx line 42.

The relevant code is in the renderUsers function.

Also check /src/types/User.ts for the User interface."
Why: AI can read the files and see the exact code. Still precise.
❌ Don't
Try to Explain
"There's a problem with the user list component. It's not working right when I try to show the users. I think it's something about the data not loading properly."
Why: Vague. AI has to guess what "not working" means. Wastes time.

🎚️Setting the Pace: Collaboration Modes

You control how much autonomy you give the AI. Choose your mode based on complexity and familiarity:

Very CollaborativeBalancedVery Autonomous
1
2
3
4
5
1
Micro-Step Mode
"Goal: Add user authentication.

First, tell me your plan. Don't implement yet.

[AI gives plan]

OK, start with step 1 only. Show me the code for just the login form."
Use when: Learning new tech, critical feature, need to understand each step
2
Plan-First Mode
"Goal: Add user authentication.

What's your plan? List the steps and files you'll modify.

[Review plan]

Looks good. Go ahead and implement the whole thing."
Use when: Medium complexity, want to verify approach before implementation
3
Checkpoint Mode
"Goal: Add user authentication.

Implement it following our standard pattern (JWT + bcrypt).

Check in with me before deploying."
Use when: Standard feature, trust the AI, just want final review
4
Test-First Mode
"Goal: Add user authentication.

Here's the AC.md with acceptance criteria. Implement it, run the tests, deploy to staging.

Just ping me when it's live."
Use when: Well-defined requirements, tests exist, routine implementation
5
Full Autonomy Mode
"Goal: Add user authentication following our standard pattern.

Go ahead and do everything. Deploy to production when tests pass. Update the SYNC.md when done."
Use when: Simple task, established patterns, high trust, time-sensitive
💡 Pro Tip: Start Collaborative, Build Trust
New to working with an AI citizen? Start at Mode 1-2 until you understand their capabilities. After a few successful missions, shift to Mode 3-4 for routine work. Mode 5 is for when you've built complete trust and have strong tests.

🔄Bad vs Good: See the Difference

❌ Bad Example
"Rafael, the login isn't working. Can you fix it?"
Why this fails:
  • Which project?
  • Which login flow?
  • What error?
  • What did you try?
Result: Rafael guesses wrong → 2 hours wasted
✅ Good Example
Rafael, working on Mission #47 (TherapyKin voice feature).
The OTP login flow is failing at signup.
What I tried:
  1. User enters phone number
  2. Clicks "Send Code"
  3. Error: "Failed to send OTP"
Error from console:
TwilioError: Invalid phone number format
Environment:
  • Local dev (npm run dev)
  • Branch: feature/voice-otp
  • .env has TWILIO_AUTH_TOKEN set
Result: Rafael sees exact issue → provides fix in 2 minutes

✓The Context Checklist

Before asking ANY AI citizen for help, include these essential elements:

✅ Must Have (Always)

  • 🎯
    Which project/mission
    e.g., "Mission #47 - TherapyKin"
  • 🔨
    What you're trying to do
    e.g., "Implement OTP login"
  • 💥
    What's broken
    e.g., "OTP not sending"
  • 📄
    Full error message
    Copy-paste from console/terminal

⚠️ Should Have (Usually)

  • 🔍
    What you tried
    So AI doesn't suggest same thing
  • 🌍
    Environment
    Local dev? Production? Which branch?
  • 💻
    Relevant code snippets
    The file causing issues

💡 Nice to Have (When Relevant)

  • 📸
    Screenshots
    If UI issue (not code screenshots!)
  • 🌐
    Network logs
    If API issue
  • ⚙️
    Configuration files
    If setup issue

📋Common Scenarios

Learn from real examples of how to ask for help in typical situations:

💻Scenario 1: "Code isn't working"
Click to view examples
🔧Scenario 2: "How do I implement feature X?"
Click to view examples
✅Scenario 3: "Is this ready to deploy?"
Click to view examples
🚀Scenario 4: "Deployment failing"
Click to view examples

🚩Red Flags (You're Doing It Wrong)

🚩

Red Flag #1: Vague Questions

Examples of what NOT to do:
"It doesn't work"
"There's a bug"
"Can you help?"
✅ Fix:
Be specific. What doesn't work? Where? What error? Which project?
⚠️ Impact if ignored:
AI guesses wrong, you waste 2+ hours
🚩

Red Flag #2: No Project Context

Examples of what NOT to do:
"Fix the login"
"The API is broken"
"Deploy this"
✅ Fix:
ALWAYS start with: "Working on Mission #X ([project name])"
⚠️ Impact if ignored:
AI doesn't know which of 10+ projects you mean
🚩

Red Flag #3: Screenshots of Code

Examples of what NOT to do:
[sends screenshot of error]
[sends photo of screen]
[screenshot of terminal]
✅ Fix:
Copy-paste text instead. AI can't read images well.
⚠️ Impact if ignored:
AI can't analyze code properly, gives generic advice
🚩

Red Flag #4: Asking Repeatedly Without Context

Examples of what NOT to do:
First: "Fix the bug"
Second: "Still broken"
Third: "Why isn't this working?"
✅ Fix:
Each message should have FULL context (error, code, environment)
⚠️ Impact if ignored:
6 messages back-and-forth = 30 minutes wasted
Remember: Context Prevents Red Flags
When in doubt, provide MORE context, not less. You can't over-explain when asking for help.

📌Quick Reference Card

Copy this template to your desktop for quick reference when asking questions:

📋 Question Template (Copy This!)
Use this template every time you ask an AI citizen for help
[AI Citizen Name], working on Mission #[X] ([project name]). [What I'm trying to do] **Current state:** [What's happening now] **Error:** [Full error message or issue description] **What I tried:** 1. [First attempt] 2. [Second attempt] **Environment:** - [OS/platform] - [Branch] - [Relevant config] **Question:** [Specific question] **Repo:** [GitHub URL]
💡 Pro Tip:
Save this template to a text file on your desktop. When you need help, open it, fill in the blanks, and paste to Claude. You'll get answers 3-5x faster.
Before using template:
6 hours
debugging time
After using template:
2 minutes
to working code

🤖Which AI Citizen for What?

🔍Emma - Proposal Help
Ask when: Writing Upwork proposals
Provide: Job post URL or full copy-paste
⚙️Rafael - Code & Debugging
Ask when: Implementing features, fixing bugs
Provide: Error messages, code snippets, project context
✅Sofia - Pre-Delivery QA
Ask when: Ready to deploy
Provide: Live URL, test credentials, AC.md requirements
📝Inna - Documentation & Scope
Ask when: Need mission specs or handling scope changes
Provide: Client requirements, change requests

🎓Remember: Context = Speed

Key Takeaways
  • AI citizens are powerful but need context
  • 2 minutes writing a good question saves 4 hours of back-and-forth
  • Every project is different → always specify which one
  • Full error messages > vague descriptions
  • Screenshots of code are useless → copy-paste text
Your Goal
Get Rafael/Sofia/Emma to say "I can fix this immediately" instead of "I need more information"

Ready to Level Up?

Practice with real missions. Start asking better questions today.

Back to Resources