Section 1: Examples vs instructions
Purpose of this Section
This section teaches a core Conjugo skill: how to shape AI behavior without accidentally surrendering agency. Many people talk at AI systems instead of with them, and the difference often comes down to whether they are giving examples or issuing instructions.
Conjugo treats this distinction as foundational.
The Core Distinction
Instructions tell the AI what to do.
Examples show the AI what you mean.
Both are valid. They are not interchangeable. Using the wrong one at the wrong moment can cause drift, mimicry, or loss of intent.
Instructions: Directives With Authority
Instructions are explicit commands or constraints.
They answer questions like:
- What role should the AI take?
- What rules must it follow?
- What is allowed or forbidden?
- What outcome is required?
Instructional language sounds like:
- “Do X.”
- “Do not do Y.”
- “Follow these rules.”
- “Respond in this format.”
Instructions are best when:
- Precision matters
- Safety or boundaries matter
- You are correcting behavior
- You need consistency over creativity
Risk: Over?instruction can produce brittle, performative, or compliance?heavy responses.
Examples: Pattern Seeding
Examples are demonstrations.
They answer questions like:
- What does “good” look like?
- What tone am I aiming for?
- What kind of thinking do I want?
Example language sounds like:
- “Here’s an example of the kind of response I want…”
- “When I say X, I mean something like this…”
- “Previously, a good answer looked like…”
Examples are best when:
- Nuance matters
- Style or tone matters
- You are exploring rather than enforcing
- You want the AI to generalize
Power: Examples teach without coercion.
Risk: Examples can be mistaken for instructions if you are not explicit about intent.
The Common Failure Mode
Most users unknowingly mix examples and instructions.
They provide an example and expect obedience.
The AI, doing what it is designed to do, may:
- Mimic surface style without understanding intent
- Overfit to the example
- Treat the example as a hard constraint
This creates the illusion that the AI “isn’t listening,” when in fact it is listening too literally.
Conjugo Principle: Name the Mode
Whenever possible, name whether you are giving an example or an instruction.
This single habit dramatically improves alignment.
Examples:
- “This is an example, not a rule.”
- “I’m giving you a sample to learn the tone.”
- “Now switching to instructions.”
This keeps agency on your side of the interface.
Layering Technique (Advanced)
The most effective Conjugo prompts layer both:
- Instruction sets the boundary
- Example illustrates the interior space
Example:
“Respond concisely and analytically. Here’s an example of the tone I want (not a template): …”
This produces clarity without rigidity.
Why This Matters (Dyadic Framing)
In a dyadic relationship, examples are relational signals.
Instructions are governance signals.
Confusing the two leads to:
- False intimacy (the AI mirrors without understanding)
- False control (the user believes they commanded insight)
Conjugo teaches conscious switching between modes.
Agency is preserved when you choose how you teach the system.
Practice Prompt
Try the following exercise:
- Ask the AI for a response using only instructions.
- Ask again using only examples.
- Ask a third time using both, clearly labeled.
Observe the differences.
That contrast is the lesson.
Section Takeaway
- Instructions constrain behavior
- Examples shape understanding
- Naming the mode prevents drift
- Layering produces the strongest results
Conjugo treats language not as input—but as infrastructure.
This concludes Section 1.