Module 1 · The Perception Audit

Exercise

The Perception Audit

Overview

This is the foundational exercise of the entire course. Every module that follows builds on what you design here. Take it seriously. Not because getting it wrong would be catastrophic (you’ll refine it as you learn more), but because the thinking you do here trains the architecture mindset. You’re not writing a list. You’re designing the perceptual structure of a system.

Instructions

Step 1: Choose your domain.

Pick a domain you know well. This should be a field where you have real experience. Your job, your business, a professional skill, an area where you’ve spent enough time to know what’s hard, what’s important, and where things go wrong. The course works best when you design for a domain you actually care about, because real knowledge reveals design insights that hypothetical domains can’t.

Good domain choices:

Your actual job (marketing team, engineering team, consulting practice, research group)
A business you’re building or planning
A professional workflow you manage (content creation, client acquisition, product development)
A domain you consult in or would like to consult in

Less effective domain choices (still workable):

A hobby you know deeply (could work if complex enough, like competitive gaming or event planning)
A domain you’re interested in but don’t have experience with (you’ll miss the nuances that make agents valuable)

Step 2: Identify five distinct perception types.

This is the core of the exercise. You need to find five fundamentally different ways of seeing your domain. Not five tasks. Not five job titles. Five different types of perception, five different ways of interpreting the same reality that don’t overlap.

The test for distinct: if two of your perception types would notice the same things, they’re not distinct enough. Each one should see something the others are blind to.

To help you think about this, here are some universal perception categories. Not all will apply to your domain, and your domain may have types that aren’t on this list:

Landscape perception: What does the external environment look like? Competition, trends, opportunities, threats. Seeing the terrain.
Analytical perception: What do the numbers say? Data, patterns, metrics, quantitative truth. Seeing the math.
Risk perception: What could go wrong? Failure modes, vulnerabilities, worst-case scenarios. Seeing the danger.
Quality perception: Is this good enough? Standards, craft, precision, gaps between intention and execution. Seeing the flaws.
Strategic perception: Does this fit? Alignment, coherence, trade-offs, long-term positioning. Seeing the whole.
Human perception: What do the people involved actually experience? User behavior, team dynamics, stakeholder concerns. Seeing the humans.
Temporal perception: What’s the pace? Deadlines, velocity, bottlenecks, momentum. Seeing the rhythm.
Creative perception: What could this become? Possibilities, innovations, alternatives nobody has considered. Seeing what doesn’t exist yet.
Adversarial perception: How would someone exploit this? Weaknesses, attack surfaces, the perspective of the one who wants you to fail. Seeing from the enemy’s position.
Operational perception: Does this actually work in practice? Implementation reality, process friction, the gap between plan and execution. Seeing the ground truth.

Step 3: For each perception type, write a description.

For each of your five types, write:

Name: A descriptive name (e.g., “Market Landscape Perception” or “Quality Assurance Perception”)
What it uniquely sees: One sentence describing what this perceiver notices that the others miss.
What it’s blind to: One sentence describing what this perceiver cannot see. This is equally important. It’s why you need the other four.
A scenario: A brief example from your domain where this perception type would catch something critical that a generalist would miss.

Step 4: Test for completeness.

Read through your five perception types and ask:

Is there a dimension of your domain that none of your five types would catch? If yes, you’re missing a perception type. Either add it or reconsider which five are most important.
Do any two types overlap significantly? If yes, merge them or sharpen the distinction.
If you gave all five perceptions the same scenario, would they each produce a different analysis? If two would produce the same analysis, they’re not distinct.

Deliverable

Perception Map Template

PERCEPTION MAP: [Your Domain] Domain: [Name and brief description] 1. [Perception Type Name] - Sees: [what it uniquely notices] - Blind to: [what it cannot see] - Example: [scenario where this catches what a generalist misses] 2. [Perception Type Name] - Sees: [what it uniquely notices] - Blind to: [what it cannot see] - Example: [scenario where this catches what a generalist misses] 3. [Perception Type Name] - Sees: [what it uniquely notices] - Blind to: [what it cannot see] - Example: [scenario where this catches what a generalist misses] 4. [Perception Type Name] - Sees: [what it uniquely notices] - Blind to: [what it cannot see] - Example: [scenario where this catches what a generalist misses] 5. [Perception Type Name] - Sees: [what it uniquely notices] - Blind to: [what it cannot see] - Example: [scenario where this catches what a generalist misses] COMPLETENESS CHECK: - Missing dimension? [Yes/No. If yes, what?] - Overlapping types? [Yes/No. If yes, which ones?] - Different analyses from same scenario? [Yes/No]

Self-Review Questions

These are not optional reflection prompts. They are the quality gate for the exercise. Do not move to the next module until you can answer each one honestly. If an answer reveals a weakness, revise the exercise output before continuing.

Before you consider this exercise complete, answer honestly:

Are your five types genuinely different? Give all five the same scenario, a real one from your domain. Would each one focus on something different? If two agents would produce essentially the same analysis, they’re not distinct enough.

Are you thinking in perceptions, not tasks? A task is “write the report.” A perception is “see whether the data supports the conclusion.” If your types are task-based (researcher, writer, reviewer), reframe them as perception-based. What does each one see that the others don’t?

Did you include what each type is blind to? This is the most important part. The reason you need five types is that each one has blind spots. If you can’t articulate the blind spot, you haven’t understood the perception well enough.

Could a generalist cover all five? If yes, ask yourself whether the generalist would cover each one as deeply as the specialist. If the answer is “probably not, but close enough,” this course will show you why “close enough” compounds into “not even close” across a multi-step pipeline.

Does your map reflect your domain expertise? The best perception maps come from people who know their domain deeply enough to identify the subtle, non-obvious types. If your map could apply to any domain, it’s too generic. The specificity of your domain knowledge should show.

What Happens Next

Your Perception Map becomes the foundation for everything that follows. In Module 2, you’ll take one of these perception types and design a complete agent identity for it. In Module 3, you’ll give it cognitive protocols. By Module 5, you’ll have a complete, working agent design. By Module 8, you’ll have connected multiple agents from this map into a system.

Don’t overthink the map. It will evolve. Your v1.0 Perception Map will look different from your final version, and that’s exactly the point. The course teaches evolution through operational experience. Your map is your first v1.0. It starts here. It grows with everything you learn.