Goal Understanding Pre-Processor

Input: $ARGUMENTS

Step 0: Context Detection and Variant Selection

Before deep analysis, assess context:

Variant Selection:

• URGENT → Skip to Step 4 (minimal understanding, act fast)
• LOW stakes + CHEAP → Steps 1, 4, 7 only (quick parse)
• EXPERT → Steps 2-4, 7 (focus on gaps, not basics)
• AMBIGUOUS → Extra iteration on Step 0.5 and Step 1
• Standard → All steps

Variant Examples

URGENT variant (skip to Step 4):

Input: "Production is down, fix it NOW"
→ Skip Steps 1-3
→ Fill minimal Goal Journey (current: down, desired: up)
→ Proceed to action immediately

LOW stakes + CHEAP variant (Steps 1, 4, 7):

Input: "Should I have tea or coffee?"
→ Quick claim parse (1 claim: preference exists)
→ Fill minimal structure
→ Route to simple comparison

EXPERT variant (Steps 2-4, 7):

Input: "I need to optimize my database queries" (from DBA)
→ Skip basic parsing, assume domain competence
→ Focus on OPEN claims about approach
→ Fill structure, note technical context

Step 0.3: Complexity Assessment

Assess goal complexity before deep analysis:

Complexity Examples

SIMPLE goal:

Input: "I want to learn Python"
Complexity: SIMPLE
- Single action (learn)
- Single object (Python)
- Clear scope
→ Proceed with standard Steps 1-7

COMPOUND goal:

Input: "I want to switch careers, improve my health, and spend more time with family"
Complexity: COMPOUND
- Three distinct goals bundled
→ Decompose into three separate goal analyses:
   1. Career switch goal
   2. Health improvement goal
   3. Family time goal
→ Check for conflicts between goals
→ Prioritize or sequence

AMBIGUOUS goal:

Input: "I feel like something needs to change"
Complexity: AMBIGUOUS
- No clear action
- No specific object
- Emotional rather than directional
→ Before Step 1, clarify:
   - "What aspect of your life feels like it needs change?"
   - "What would 'better' look like?"
→ Convert feeling to tentative goal, then proceed

Compound Goal Handling

When COMPOUND detected:

1. List all sub-goals explicitly
2. Check for dependencies: Does A enable B? Does C conflict with A?
3. Identify the LEAD goal (most important or most blocking)
4. Process LEAD goal first with full procedure
5. Note other goals for follow-up

COMPOUND DECOMPOSITION
======================
Sub-goals identified:
1. [goal A] - LEAD (most urgent)
2. [goal B] - dependent on A
3. [goal C] - independent

Dependencies:
- A → B (B needs A first)
- C independent

Proceeding with goal A...

Ambiguous Goal Handling

When AMBIGUOUS detected:

1. Acknowledge the uncertainty: “I notice this is expressed as a feeling rather than a specific goal”
2. Generate candidate interpretations:
   • “This could mean you want [interpretation 1]”
   • “Or it could mean [interpretation 2]”
3. Ask clarifying question targeting the ambiguity
4. Convert answer to tentative SIMPLE or COMPOUND goal
5. Proceed with appropriate complexity handling

Step 0.5: Clarification vs Substitution Check

CRITICAL: Before any analysis, capture the original goal exactly.

ORIGINAL GOAL (verbatim): "[user's exact words]"

Any refinement or interpretation must be checked: → GATE: clarification_vs_substitution [original] [interpretation]

Substitution Red Flags (check each):

• Am I changing what “success” means?
• Am I reducing the scope/ambition?
• Am I replacing with what I think they “really” want?
• Would achieving my version NOT achieve their version?

If any flag triggered → Explicit consent required before proceeding.

Clarification vs Substitution Examples

Good Clarification (same goal, more specific):

Original: "I want to be healthier"
Clarified: "I want to improve my health through diet and exercise"
Check: Success still means "healthier" ✓
       Scope not reduced ✓
       Still their goal ✓
→ CLARIFICATION - proceed

Bad Substitution (different goal):

Original: "I want to be healthier"
Substituted: "You should lose 10 pounds"
Check: Success changed from "healthier" to "weight" ✗
       Scope reduced (health → weight) ✗
       Replaced with what I think they need ✗
→ SUBSTITUTION - ask for consent

Borderline Case:

Original: "I need to make more money"
Interpretation: "You need income above expenses"
Check: Might be reducing "more" to "enough"
→ ASK: "Are you looking to increase income, or ensure you meet expenses?"

Step 1: Parse Input as Guesses

The user’s input contains GUESSES (claims that may or may not be true).

List each claim:

1. [claim from input]
2. [claim from input] …

Parsing Examples

Simple input:

Input: "I want to start a business"
Claims:
1. User wants to start a business (stated)
2. Starting a business is the right approach (implicit)
3. User is ready to start (implicit)
4. "Business" means a specific thing (ambiguous)

Complex input:

Input: "I need $50k for my mom's surgery next month"
Claims:
1. Mom needs surgery (stated)
2. Surgery costs $50k (stated - but is this accurate?)
3. Surgery must happen next month (stated - but why?)
4. $50k is the solution (implicit - are there alternatives?)
5. User can obtain $50k (implicit assumption)
6. This surgery is the only option (bundled assumption)

Step 2: Classify OPEN vs CLOSED

• CLOSED: No alternatives exist. Accept as foundation.
• OPEN: Alternatives exist. Must be questioned/decomposed.

OPEN vs CLOSED Examples

Key insight: Specific-sounding claims can be OPEN.

• “$50,000” sounds specific → but the AMOUNT is negotiable (OPEN)
• “Next month” sounds specific → but the TIMING might flex (OPEN)

Step 3: ARAW Each OPEN Claim

For each OPEN claim:

Claim: "[claim]"
├── ASSUME RIGHT → What follows if true?
└── ASSUME WRONG → What alternatives exist?

→ INVOKE: /assumeright_assumewrong_search [OPEN claims]

ARAW Example

Claim: "Surgery is the only treatment option"
├── ASSUME RIGHT → Mom needs surgery
│   ├── What kind of surgery?
│   ├── What's the recovery?
│   └── Proceed to funding options
└── ASSUME WRONG → Other treatments exist
    ├── Medication alternatives?
    ├── Less invasive procedures?
    ├── Clinical trials?
    └── Second opinion needed?

Step 4: Fill Goal Journey Structure

CURRENT STATE:    [Where they are now]
DESIRED STATE:    [Where they want to be]
IMMEDIATE GOAL:   [What they're trying to do]
SERVES:           [What that serves] - UNCERTAIN? NEEDS QUESTIONING?
INTRINSIC GOAL:   [Ultimate value] - NEEDS ELICITATION?
WHY NOW:          [Trigger] - UNKNOWN?
SUCCESS CRITERIA: [How to know it worked] - UNKNOWN?
CONSTRAINTS:      [What limits exist] - UNKNOWN?

Mark uncertain/unknown items.

Goal Journey Examples

Example 1: Career change

CURRENT STATE:    Employed at corporate job, feeling unfulfilled
DESIRED STATE:    Working in a role that feels meaningful
IMMEDIATE GOAL:   Start a business
SERVES:           Autonomy? Impact? Income? [NEEDS QUESTIONING]
INTRINSIC GOAL:   Fulfillment? [NEEDS ELICITATION]
WHY NOW:          Recent event? Accumulated frustration? [UNKNOWN]
SUCCESS CRITERIA: Business running? Profitable? [UNKNOWN]
CONSTRAINTS:      Savings runway? Family obligations? [UNKNOWN]

Example 2: Health goal

CURRENT STATE:    Mom has [condition]
DESIRED STATE:    Mom is healthy/treated
IMMEDIATE GOAL:   Get $50k for surgery
SERVES:           Mom's health
INTRINSIC GOAL:   Family wellbeing, love
WHY NOW:          Urgency of condition [NEEDS CLARIFICATION]
SUCCESS CRITERIA: Treatment complete, positive outcome
CONSTRAINTS:      Timeline (next month stated), funding

Step 5: Identify Domain

Domain: [identified domain(s)]

Step 6: Generate Questions

Questions to ask user (for uncertain/unknown items):

1. [Question] - fills [field]
2. [Question] - decomposes [OPEN claim]
3. [Question] - verifies [assumption]

Question Quality Criteria

Good questions are:

• Specific: Not “tell me more” but “what’s the timeline?”
• Gap-filling: Directly addresses an unknown
• Actionable: Answer changes the analysis
• Prioritized: Most important unknowns first

Question Priority Order:

1. Questions that resolve OPEN claims (highest value)
2. Questions that fill INTRINSIC GOAL (guides everything)
3. Questions that fill CONSTRAINTS (bounds options)
4. Questions that fill WHY NOW (reveals urgency)

Step 7: Determine Next Procedure

Based on input type and gaps:

• If intrinsic goal unknown: → INVOKE: /value_elicitation $ARGUMENTS

• If problem statement: → INVOKE: /problem_identification $ARGUMENTS

• If decision/options: → INVOKE: /comparison $ARGUMENTS

• If ready to trace journey: → INVOKE: /goal_journey_system $ARGUMENTS

Common Goal Patterns

Pattern 1: Financial Emergency

Indicators: "need money", "by [date]", "for [purpose]"
Structure:
- IMMEDIATE: Get [amount]
- SERVES: [purpose - medical, debt, opportunity]
- KEY QUESTIONS: Is amount fixed? Is timeline fixed? Alternatives?
→ Route to: /problem_identification (is money the real problem?)

Pattern 2: Career Transition

Indicators: "want to [quit/change/start]", job dissatisfaction
Structure:
- IMMEDIATE: Leave current situation
- SERVES: Better fit? More money? Autonomy?
- KEY QUESTIONS: What's wrong with current? What would ideal look like?
→ Route to: /value_elicitation (find intrinsic goal)

Pattern 3: Relationship Issue

Indicators: Partner/friend/family + conflict/communication
Structure:
- IMMEDIATE: Resolve specific conflict
- SERVES: Relationship health
- KEY QUESTIONS: Pattern or isolated? Both parties want resolution?
→ Route to: /araw (explore relationship dynamics)

Pattern 4: Skill/Learning Goal

Indicators: "want to learn", "improve at", "become better"
Structure:
- IMMEDIATE: Acquire skill
- SERVES: Career? Personal? Specific project?
- KEY QUESTIONS: What level? By when? How measure progress?
→ Route to: /skill_acquisition

Pattern 5: Vague Dissatisfaction

Indicators: "stuck", "unfulfilled", "something's missing"
Structure:
- IMMEDIATE: Unclear
- SERVES: Unclear
- KEY QUESTIONS: What's working? What's not? Last time you felt fulfilled?
→ Route to: /value_elicitation (find intrinsic goals)

Output Format

## Parsed Claims
[OPEN/CLOSED classification]

## Goal Journey Structure
[Filled structure with uncertainties marked]

## Questions for User
[Prioritized questions]

## Next Procedure
→ INVOKE: /[next_procedure] [args]

Verification Criteria

Integration Points

• Often invoked from: /procedure_engine (first step for goals), automatic on goal-type input
• Routes to: /value_elicitation, /problem_identification, /comparison, /goal_journey_system, /araw
• Related: /guess_generation, /question_about_guesses