Learn → Practice Studio → Lab 3
Lab 3 Intermediate Judgment

Context Overload

Learn when more context helps — and when it hurts. Master the art of including only what's relevant.

~30 minutes
4 Exercises
Builds on Labs 1-2
Begin Learning Back to Practice Studio
The Balancing Act
Finding the Right Amount
Too Little

AI guesses, invents details

Just Right

Focused, relevant, complete

Too Much

Signal drowns in noise

More context ≠ better results

What You'll Learn

1

Understand why excessive context degrades AI performance

2

Distinguish between signal (relevant) and noise (extraneous) in prompts

3

Apply the "necessary and sufficient" principle to context selection

4

Practice trimming and expanding context for optimal results

The Core Concept

The Paradox of Context

Your intuition says: "Give the AI everything it might need. More information is better." This intuition is wrong — and understanding why is one of the most valuable skills in prompt engineering.

AI attention is a finite resource. Every piece of context you provide competes for that attention. Irrelevant details don't just waste space — they actively dilute focus on what matters. Your actual requirements get buried under a pile of "just in case" information.

Too little context means the AI guesses and fills gaps, often incorrectly. Too much context means the signal drowns in noise. The goal isn't maximum context — it's optimal context. Everything necessary, nothing more.

Overloaded
"Hi! I hope you're doing well today. I'm working on a project for my company (we're a mid-sized tech firm founded in 2019, about 200 employees). Anyway, I need help writing an email. The email is for our customers. We want to announce a new feature. The feature is called SmartSync. It's basically a way to sync data automatically. Can you help? Thanks so much!"
⚠ Core request buried under pleasantries, backstory, hedging
vs
Calibrated
"Write a customer announcement email for SmartSync, a feature that automatically synchronizes data between devices. Audience: B2B tech professionals. Tone: Professional but friendly. Length: 150-200 words. Include: feature benefits, availability date (March 1), and upgrade path."
✓ All necessary context, zero noise, clear requirements
Key Insight: The AI's attention is a finite resource. Every irrelevant detail competes with your actual requirements.
Your Toolkit

The N.E.S. Principle

For every piece of context, ask these three questions to calibrate your prompt.

N

Necessary

Would the output change meaningfully if I removed this?

E

Eliminable

What background have I included that doesn't affect the output?

S

Sufficient

Is anything critical still missing that forces the AI to guess?

Remember the Principle
Necessary → Eliminable → Sufficient

Everything needed, nothing more.

The Context Test

"Would the output change meaningfully if I removed this?"
Yes → Keep it
(Necessary)
No → Remove it
(Eliminable)
"Is anything critical still missing?"
Yes → Add it
(Sufficient)
No → Done
(Calibrated)
Key Distinction

Separating What Matters

Learn to recognize signal (keep) versus noise (remove) in your prompts.

Signal (Keep)

  • Task requirements — what you want produced
  • Output format specifications
  • Constraints that affect the result
  • Relevant domain context
  • Success criteria

Noise (Remove)

  • Your thought process — "I was thinking maybe..."
  • Unnecessary backstory — "Last Tuesday, I..."
  • Redundant instructions — same thing, three ways
  • Pleasantries — "Could you please kindly..."
  • Unrelated context — "By the way, we also..."

Same Request, Different Clarity

Cluttered

"Hi there! I hope you're having a great day. So basically, I'm working on a research paper for my graduate program — I'm studying at Mumbai University, by the way, which is pretty demanding. I need to find some statistics about renewable energy adoption in India. I've been searching for a while and haven't found what I need. Could you maybe possibly help me find some good data? Thanks so much in advance!"

Clean

"Provide recent statistics on renewable energy adoption in India. Include: solar and wind capacity growth (2020-2024), government targets, and investment figures. Cite sources. Format as a bulleted list with years and figures."

Practice Exercise

Your Turn: Calibrate the Context

Each scenario presents a prompt with context issues. Identify the problems, then reveal the analysis to check your thinking.

Business Trim the Bloat
Scenario 1 of 4
Original Prompt
"Hi! I hope you're doing well. I was wondering if you could help me with something. So basically, our company, which is a mid-sized B2B SaaS company in the HR tech space (we've been around since 2019 and have about 200 employees now), needs to send out an email to our customers. The email is for marketing purposes. We want to announce our new feature. The feature is called 'Smart Scheduling' and it helps HR managers schedule interviews more efficiently. We want the email to be professional but also friendly. Not too formal, you know? But also not too casual. Maybe around 150-200 words? Thanks in advance for your help!"
Your Task

Identify what's signal (necessary) and what's noise (eliminable) in this prompt.

Context Analysis
✕ Remove: "Hi! I hope you're doing well" — pleasantry, no impact
✕ Remove: "I was wondering if you could help me" — hedging
✕ Remove: Company founding year and employee count — irrelevant to email content
✓ Keep: B2B SaaS, HR tech — defines audience context
✓ Keep: Feature name and description
✓ Keep: Tone and length requirements
Calibrated Version

"Write a marketing email announcing 'Smart Scheduling' — a feature that helps HR managers schedule interviews efficiently. Audience: HR professionals at B2B companies. Tone: Professional but approachable. Length: 150-200 words."

Technical Fill the Gaps
Scenario 2 of 4
Original Prompt
"Fix my code."
Your Task

Identify what critical context is missing that would force the AI to guess.

Context Analysis
+ Missing: What programming language?
+ Missing: What is the code supposed to do?
+ Missing: What's the actual error or problem?
+ Missing: The code itself
+ Optional but helpful: What have you already tried?
Sufficient Version

"Fix this Python function that should return the sum of even numbers in a list. Currently returns 0 for all inputs. Here's the code: [code block]. I think the issue is in the condition, but I'm not sure."

Research Find the Signal
Scenario 3 of 4
Original Prompt
"I'm working on a research paper about climate change adaptation in coastal cities. I've been researching this topic for about three months now as part of my master's thesis at Mumbai University. My advisor suggested I look at case studies from Southeast Asia. Specifically, I need to understand how cities like Jakarta, Bangkok, and Ho Chi Minh City are implementing flood management systems. I want you to compare their approaches, focusing on infrastructure investment, policy frameworks, and community engagement. The paper is due next month so I'm a bit stressed. Please format as a structured comparison with clear categories. About 500 words per city would be ideal."
Your Task

Separate the signal (relevant context) from the noise (extraneous details).

Context Analysis
✓ Signal: Topic — climate adaptation in coastal cities
✓ Signal: Cities — Jakarta, Bangkok, Ho Chi Minh City
✓ Signal: Focus areas — infrastructure, policy, community engagement
✓ Signal: Format — structured comparison, 500 words/city
✕ Noise: Duration of research (3 months)
✕ Noise: University and advisor details
✕ Noise: Personal stress about deadline
Calibrated Version

"Compare flood management approaches in Jakarta, Bangkok, and Ho Chi Minh City. Focus areas: infrastructure investment, policy frameworks, and community engagement. Format: Structured comparison with clear categories. Length: ~500 words per city."

Real World Full Calibration
Scenario 4 of 4
Original Prompt
"So I'm a product manager and we have this quarterly review coming up next week. I need to present to the leadership team (CEO, CFO, VP of Engineering, VP of Sales). We launched three features this quarter: (1) the new dashboard redesign that took forever but finally shipped, (2) API v2 which the engineering team is really proud of, and (3) mobile push notifications. Adoption has been mixed — dashboard is doing great, API is slow (enterprise sales cycle), mobile is underperforming. I want to write a presentation that shows we're making progress without hiding the problems. My last quarterly review didn't go great because I was too optimistic. Help me write talking points that are honest but also show the value we're delivering. Oh and keep it to maybe 3-4 bullet points per feature max."
Your Task

Rewrite this prompt with optimal context — keeping what matters, removing what doesn't.

Full Calibration Analysis
✓ Keep: Three features and their adoption status
✓ Keep: Audience (executive leadership)
✓ Keep: Tone requirement (honest but value-focused)
✓ Keep: Format constraint (3-4 bullets per feature)
✕ Remove: "dashboard took forever" — editorial, not relevant
✕ Remove: "engineering team is really proud" — internal sentiment
✕ Remove: Last review performance — past context, not task-relevant
Calibrated Version

"Create quarterly review talking points for three product launches:

1. Dashboard redesign — strong adoption
2. API v2 — slow uptake (enterprise sales cycle)
3. Mobile notifications — underperforming

Audience: Executive leadership (CEO, CFO, VPs)
Tone: Honest about challenges while highlighting progress
Format: 3-4 bullet points per feature
Focus: Value delivered and realistic next steps"

Watch For

Context Anti-Patterns

These common habits undermine prompt effectiveness. Learn to recognize them in your own writing.

The Kitchen Sink

Including everything "just in case" the AI might need it.

Example

"Here's our company history, org chart, product roadmap, financial statements, competitor analysis, and customer feedback from the last three years. Now write a tweet."

Fix: Ask "Would removing this change the output?" If no, remove it.

The Life Story

Extensive backstory unrelated to the actual task.

Example

"I've been working in marketing for 15 years. Started at a small agency, moved to corporate, had a brief stint in consulting... Anyway, can you proofread this email?"

Fix: Lead with the task, add context only if it affects the output.

The Hedged Request

So many qualifiers that the actual ask gets buried.

Example

"I was wondering if maybe you could possibly help me with something, if it's not too much trouble, and only if you think it makes sense, but feel free to push back..."

Fix: State your request directly. Qualifiers rarely improve output.

The Duplicate

Same instruction phrased three different ways.

Example

"Make it concise. Keep it brief. Don't make it too long. Aim for short and punchy. Brevity is key. No unnecessary words."

Fix: Say it once, clearly. Repetition adds noise, not emphasis.
Reflect

What Did You Learn?

Guided Reflection

Take a moment to consider these questions:

  • 1 Think of a recent prompt you wrote. What context could you have removed?
  • 2 When might you intentionally include MORE context than strictly necessary?
  • 3 How does this lab change your approach to drafting prompts?

Key Takeaways

  • Context is about relevance, not volume — more is not better
  • Every piece of context should pass the "Would output change?" test
  • The N.E.S. principle: Necessary, Eliminable, Sufficient
Your Learning Journey

The Bloom Journey

You're in the "Grow" stage — building foundational skills through deliberate practice.

Seed
Discovery
Grow
Foundations
Flourish
Application
Thrive
Mastery
Radiant
Leadership
Lab 3 of 5 in Practice Studio You're developing judgment skills — learning to calibrate context for optimal AI collaboration.
Previous: Spot the Hallucination Return to Practice Studio Next: Role Prompting
Lab 3 of 5 complete

Ready for the Next Challenge?

Continue developing your judgment with Role Prompting, or review all labs.

Next: Role Prompting View All Labs