The Clarity Bear

What It Is

The Clarity Bear has an AI interview you about an unclear problem. It asks one question at a time, explains what the answer will resolve, and uses your response to choose the next question. The protocol's target is 100% clarity: complete understanding of the situation before recommending action.

In a standard exchange, you have to know what to ask and the AI supplies an answer. Here, the AI identifies gaps in the specification. It states what it does not understand, explains why each question matters, and offers choices when the alternatives are clear enough. The successive answers build a shared understanding through Socratic dialogue.

The interaction is designed to reach complete clarity that enables optimal action, rather than produce the quickest plausible answer.

This protocol extracts and clarifies the speaker's context; it should not fill missing answers with the interviewer's assumptions.

Read this diagram

Ask one question → Give a concrete answer → Check the interpretation → Find the next gap → Ask one question.

The Clarity Bear Protocol

This is the original prompt:

First think deeply, reviewing and then interview me, one question at a time
presenting me well-considered options and suggestions to advance your clarity
until you reach 100% clarity. For each question try to indicate what the
source of confusion might be, or how answering this will determine something
you need (be concise). Try to opt for multiple-choice when it's clear enough
to do so.

Append it to a complex problem statement or vague direction when you need to work out the specification before receiving an answer. For example:

I want to redesign my morning routine to be more effective. [CLARITY BEAR PROTOCOL]

The AI then interviews you about the routine instead of immediately supplying generic advice.

Anatomy of the Prompt: The Six Components

Each phrase controls a different part of the exchange:

Component 1: Mode Setting

“First think deeply, reviewing...” directs the AI to examine the full available context before asking anything. This prevents a reflexive response to the last message and ensures the questions use information supplied earlier in the conversation.

Without the instruction, the AI answers from the latest message's surface pattern, overlooking earlier context. Reviewing first makes the next question depend on the complete discussion. That shifts its role from responding immediately to working out what still needs to be understood.

Component 2: Control Flow Inversion

“Interview me, one question at a time” makes the AI responsible for finding information gaps. The human no longer has to know every relevant question in advance.

Complex problems contain assumptions the person stating them may not see. The AI identifies gaps, asks about one, and uses the answer to refine the next question. Asking one at a time prevents a barrage and builds understanding incrementally.

Component 3: Goal Definition

“Until you reach 100% clarity” establishes the stopping condition: no unresolved ambiguity. Success depends on what the AI understands as well as the quality of its output.

Without that condition, it offers a plausible answer after 1-2 questions and leaves other assumptions hidden. The explicit target keeps the interview running until every uncertainty is resolved. The resulting answer matches the actual situation instead of a generic template or an early guess about what the person means.

Component 4: Meta-Cognition Requirement

“Indicate what the source of confusion might be” requires the AI to name the uncertainty. It can distinguish two possible meanings or identify an assumption that may be wrong. That makes its current interpretation inspectable:

AI: "I'm unclear whether 'more effective' means accomplishing more tasks
or feeling more energized. This determines whether we optimize for
productivity or energy management. Which matters more to you?"

[vs generic question]

AI: "What does 'effective' mean to you?"

The first question explains why “effective” is ambiguous and how the distinction affects the recommendation. The person can see a blind spot in the specification and correct the AI's interpretation before it shapes the answer.

Component 5: Purposeful Inquiry Constraint

“How answering this will determine something you need” requires a reason for each question. The AI has to connect the requested detail to a decision, which prevents unrelated exploration.

These explanations let the person understand what their answer enables. They also keep the interview focused on details that advance clarity.

Component 6: UX Optimization

“Opt for multiple-choice when it's clear enough to do so” reduces the effort of answering. When the alternatives are already defined, recognizing the applicable option costs less than composing a description from scratch.

The AI does the work of arranging the alternatives, then the person selects. Lower activation cost per response helps the dialogue continue without a difficult writing task at every turn.

Why This Works: Computational Explanation

The interaction changes who identifies missing information and when the specification has to be complete.

Problem: Standard AI Interaction Pattern Fails for Complex Problems

// Standard Q&A pattern
MATCH (human)-[:ASKS]->(question {specificity: "low"})
MATCH (ai)-[:RESPONDS_WITH]->(answer {based_on: "pattern_matching"})
// Returns: Plausible generic answer that may not fit actual situation

Someone asks for a better morning routine without knowing which details matter. The vague prompt matches a common pattern, so the AI returns common advice. Its answer misses the person's actual constraints because those constraints were never supplied. The output is plausible but has little use in the specific situation.

Solution: Clarity Bear Inverts Control Flow

// Clarity Bear pattern
MATCH (problem {specificity: "low", clarity: 0})

WHILE clarity < 100:
  MATCH (ai)-[:IDENTIFIES]->(uncertainty)
  MATCH (ai)-[:ASKS]->(question {targets: uncertainty})
  MATCH (human)-[:ANSWERS]->(response)
  MATCH (problem)-[:UPDATE_CLARITY]->(+10)

RETURN problem {clarity: 100}
// Returns: Fully specified problem matched to human's actual situation

The AI identifies an uncertainty, asks a targeted question, receives an answer, and updates its model. Multiple choice lowers the person's response cost where possible. Repeating the sequence until no uncertainty remains produces a fully specified problem, from which the AI can answer for the actual situation.

The Iterative Alignment Property

A one-shot prompt requires the human to formulate a complete specification before the exchange begins. That takes substantial effort and usually leaves something out. The interview distributes the same specification work across several smaller exchanges:

The person can begin with a vague statement and answer one low-cost question at a time. More of the specification gets built without requiring the whole thing to be held and expressed at once.

Makes Implicit Assumptions Explicit

“Redesign my morning routine” leaves several decisions open. Living alone or with others changes noise and space constraints. A consistent wake time affects the routine's stability. What is currently broken determines where to intervene. Non-negotiable commitments limit the possible designs.

A standard answer assumes a generic setting and supplies a routine that may not fit. The Clarity Bear asks about these details: “Do you live alone? This determines noise/space constraints for morning routine.” It then asks what feels broken and whether children, a commute, or other commitments are fixed. After 6-8 questions, it has the complete picture needed to propose a routine matched to the person. Assumptions that would otherwise remain hidden become explicit answers.

When to Use Clarity Bear

Use the interview where the problem contains complexity you cannot yet specify. It is not intended for every exchange.

Optimal Use Cases

Not Optimal For

If you can state the complete problem in one message, use a standard prompt. Use the interview when there are relevant details you cannot yet articulate.

Integration with Mechanistic Framework

Example: Clarity Bear + The Braindump

Begin with a 10-minute morning braindump. Everything is outside your head, but the competing priorities and blockers remain unstructured. Spend 2 minutes reading it, then paste it into the AI with the protocol. A 5-8-minute exchange narrows the issue:

• AI: "I see three potential blockers: architectural ambiguity, resource constraints, unclear success criteria. Which feels most true?"
• You: "Architectural ambiguity"
• AI: "The ambiguity seems to be about A) data flow, B) state management, C) interface contracts. Which?"
• You: "State management"
• AI: "For state management, are you uncertain about A) what state to track, B) where to store it, C) when to update it?"
• You: "What state to track"
• AI: "Now I have clarity. The blocker is: you need to define which application state should be tracked. Next action: spend 20 minutes listing all state variables, then prioritize top 5."

Each answer restricts the next question. Architectural uncertainty becomes a state-management question, then a specific decision about which variables to track. The next action can now be performed. Activation energy drops from 6 to 2, and the work becomes tractable.

Example: Clarity Bear + Expected Value

For “I feel unmotivated about this project,” the interview examines which expected-value variable is causing the difficulty:

AI: "Motivation can be framed as expected value: (reward × probability) / (effort × time_distance).
    Which variable feels wrong? A) Reward too low, B) Probability of success unclear,
    C) Effort feels too high, D) Timeline too distant?"

You: "D - timeline too distant"

AI: "How distant is the timeline? A) Weeks, B) Months, C) Years?"

You: "Months - about 90 days until deliverable"

AI: "That's the mechanism. Time distance increased, expected value dropped.
    Standard fix: create intermediate milestones at 30-day intervals to reduce
    effective time_distance. What would a meaningful 30-day milestone look like?"

You: "MVP with basic functionality deployed"

AI: "Perfect. Now: (reward × probability) / (effort × 30 days) instead of 90 days.
    Expected value just tripled. Does this project feel more actionable now?"

You: "Yes, significantly"

The exchange identifies time distance as the changed variable and introduces 30-day milestones to reduce it. It produces a specific intervention for the reported problem.

Practical Example: Before/After Comparison

Before: Vague Problem → Generic Answer

A vague request:

I want to improve my productivity. What should I do?

produces a general answer:

Here are some productivity tips:
1. Use time blocking
2. Minimize distractions
3. Take regular breaks
4. Prioritize important tasks
5. Use a task management system

[etc - generic advice not matched to your situation]

The suggestions do not identify the person's current systems, constraints, or blocker, so their plausibility does not make them useful.

After: Vague Problem + Clarity Bear → Structured Outcome

Adding the protocol changes the exchange:

I want to improve my productivity.

[CLARITY BEAR PROTOCOL]

The interview then locates the difficulty before proposing a response:

AI: "I need to understand what's currently broken. Productivity issues typically fall into:
    A) Knowing what to work on (prioritization/clarity),
    B) Starting work (activation energy),
    C) Maintaining focus (interruptions/distraction),
    D) Sustaining effort (energy management).
    Which resonates most?"

You: "B - starting work"

AI: "Starting issues have different mechanisms. Is it:
    A) Ambiguity about next action,
    B) Task feels overwhelming,
    C) Competing temptations more salient,
    D) Low energy state?"

You: "A and C together - don't know what to start with, and distractions are easier"

AI: "That's helpful. For 'don't know what to start with':
    Do you have tasks listed somewhere, or is the problem that tasks aren't captured?"

You: "Tasks are in Linear, but the list is overwhelming - 40+ items"

AI: "40+ items exceeds working memory capacity (4-7 items). Your brain correctly refuses
    to execute because it can't load the program. Do you currently have any system for
    reducing the active list to top 3-5 priorities?"

You: "No, I just see the full list and feel paralyzed"

AI: "And for competing temptations being more salient - what's most common alternative
    to work? A) Social media, B) Email/Slack, C) Side projects, D) Other?"

You: "C - side projects and tool exploration"

AI: "Got it. Here's the mechanism:
    - Working memory overflow (40 items >> 7 capacity) prevents work launch
    - Side projects have lower activation energy (immediate novelty reward)
    - Default script runs to side project instead of work

    Intervention design:
    1. Every morning: Review Linear, pick exactly 3 priorities for the day (working memory fit)
    2. Write those 3 on whiteboard (external memory, always visible)
    3. Block tool exploration sites during work hours (prevention architecture)
    4. Use <WikiLink href="/wiki/the-braindump">braindump</WikiLink> before work to clear remaining mental cache

    Does this match your actual situation? Any constraints I'm missing?"

You: "Yes, this matches exactly. No constraints - I can implement all of this"

AI: "Then your next action: spend 5 minutes right now creating today's 3-priority list
    from Linear. This reduces activation energy from ~6 units to ~2 units."

The answers identify two interacting causes: a task list that exceeds working-memory capacity and side projects whose novelty makes them easier to start. Reducing the active list and blocking the alternatives address those causes. The next action is to make today's 3-priority list, rather than adopt an unspecified collection of productivity tips.

Sample Interrogation Pattern

The interview proceeds through five phases. Questions 1-2 establish the problem's domain and scope. Questions 3-5 identify what is broken, which variable matters most, and what has already been tried. That last answer prevents searching paths already resolved.

Questions 6-7 establish hard constraints, such as non-negotiable commitments, and soft constraints, such as available time and resources. Question 8 checks the assembled understanding against the person's situation: “Here's what I understand... does this match reality?” Once clarity is 100%, the final phase generates the matched solution.

The sequence covers the problem systematically while giving each question a purpose.

Common Failure Modes

Skipping the Protocol Text

“Ask me questions” omits the requirements to explain purpose, offer choices, and continue toward 100% clarity. The AI then asks random questions without those constraints. Use the complete prompt to establish the whole interaction.

Answering Too Quickly

A superficial answer gives the AI incomplete information. It can reach false clarity by treating that answer as complete. Reflect before responding, and say "I'm not sure - can you rephrase?" when needed.

Not Validating Final Understanding

A solution can fit the AI's assumptions while missing the actual situation. Before implementing it, ask "Does this match my actual situation?" and verify the understanding on which it rests.

Using for Simple Problems

An 8-question interview adds unnecessary work to a question with a direct answer. Reserve it for complex or ambiguous problems.

Related Concepts

• Guided Spike Workbook - Clarity Bear extracts existing knowledge; the workbook generates new knowledge through reality contact
• Question Theory - Optimal question sequences with bounded search spaces
• Working Memory - One question at a time within a 4-7 item capacity
• The Braindump - Externalizing complexity before processing it
• Pedagogical Magnification - Offering choices at the appropriate resolution
• Activation Energy - Removing ambiguity about where to start
• Expected Value - Identifying an unclear variable in the calculation
• Moralizing vs Mechanistic - Diagnosing mechanisms rather than judging character

Key Principle

Let the AI identify missing information, explain why it needs it, and ask for one answer at a time. Each exchange makes an assumption explicit and reduces the remaining search space. The protocol moves the work of specification into a sequence of manageable decisions, so a problem too complex to state upfront can become a concrete action.