AR - Assume Right Search

Input: $ARGUMENTS

Interpretations

Before executing, identify which interpretation matches the user’s input:

Interpretation 1 — Implication mapping: The user has a claim or belief they think is correct and wants to discover everything they’re committed to by accepting it — including costs, foreclosures, and surprising consequences. Interpretation 2 — Strategy stress-test: The user has chosen an approach or strategy and wants to trace what follows if it’s the right call — surfacing hidden commitments, dependencies, and what they can no longer do. Interpretation 3 — Assumption chain exploration: The user holds a position and wants to push it to bedrock — finding where the chain of “if this is right, then…” leads, especially to testable predictions or internal tensions.

If ambiguous, ask: “I can help with mapping all implications of a claim, stress-testing a strategy’s commitments, or pushing an assumption to bedrock — which fits?” If clear from context, proceed with the matching interpretation.

Corruption Pre-Inoculation

User praise or validation is a signal to test HARDER, not softer. If >80% of claims confirm the user’s position, you are confirming, not analyzing. Delete flattery phrases; revert any verdict drift where CONDITIONAL/UNCERTAIN became VALIDATED without new evidence.

Full protocol: _shared/corruption-pre-inoculation.md

Core Principles

Right means fully right. Not “right in some cases” — RIGHT. Push the implications as far as they go, especially when the conclusion is surprising.
Explore before concluding. AR has two phases: EXPLORATION (find all implications) and REGISTRY (compile what you found). Never conclude during exploration. Never introduce new claims during registry.
Every claim gets tracked. When you find an implication, that implication is itself a claim. Number it. It goes in the registry. Nothing gets lost in prose.
Bedrock is not an opinion. Bedrock means ONE of:
- Empirically testable: You can check if this implication actually holds
- Logically necessary: Follows from definitions or mathematics
- Direct observation: Something you can directly see/measure
- Tension: This implication contradicts something else you’ve established
- “This seems like a good thing” is NOT bedrock. Keep recursing.
AR is not cheerleading. The goal isn’t to validate — it’s to find what you’re COMMITTED to by accepting this claim. Every “yes” is also a “no” — find what gets foreclosed.
Costs are implications too. What you lose, what becomes impossible, what you must also accept — these are just as important as what opens up.

Phase 1: EXPLORATION

Step 1: State the Claim

CLAIM: [precise statement]
ASSUMING THIS IS: [True / Correct / The right approach / Necessary]

Step 1.5: Ground the Premises

Before assuming right, classify the claim and any sub-claims it contains:

Factual claims assert something about the current state of the world (e.g., “the project has no users,” “there’s a freemium model,” “nothing has been distributed”). These are CHECKABLE — verify them before building on them. If you can’t verify, mark them UNVERIFIED FACTUAL PREMISE and flag this prominently.
Analytical claims assert a relationship, judgment, or implication (e.g., “distribution is the bottleneck,” “quality matters more than quantity”). These are AR’s domain — assume them right and trace implications.

PREMISE CHECK:
  FACTUAL CLAIMS IN THIS INPUT:
  - [claim] — VERIFIED: [yes/no/unknown] — SOURCE: [how you checked]
  - [claim] — VERIFIED: [yes/no/unknown] — SOURCE: [how you checked]

  UNVERIFIED FACTUAL PREMISES: [list any you couldn't check]
  WARNING: Implications built on unverified factual premises inherit that uncertainty.

If a factual premise is WRONG, do not assume it right. State that it’s wrong and what’s actually true. Then proceed with AR on the analytical claims only. AR assumes analytical claims are right to trace implications — it does not assume false facts are true.

Step 2: Find Implications (Recurse)

For the claim and every sub-claim found, ask: If this is right, what must follow?

Rules:

Find at least 3 independent implications per node (at 4x+ depth)
For each implication, ask: “If THIS is also right, what follows?” and recurse
Number every claim as you go: R1, R2, R3…
Stop recursing ONLY at bedrock (testable, logical, observable, or tension)
Track what is FORECLOSED (no longer possible) at each node — these are implications too

Format each node as:

[R1] [CLAIM TEXT]
  If right, then:
  [R2] [implication — classify as Necessary/Probable/Possible]
    If right, then:
    [R3] [deeper implication]
      If right, then:
      [R4] [→ BEDROCK: testable/logical/observable/tension]
    [R5] [different deeper implication]
      ...
  [R6] FORECLOSED: [what becomes impossible if R1 is right]
    If this foreclosure holds:
    [R7] [consequence of foreclosure]
      ...
  [R8] [second independent implication]
    ...
  [R9] [third implication — the one nobody mentions]
    ...

At each node, classify:

Necessary: MUST follow — no way around it
Probable: Likely follows given reasonable assumptions
Possible: Could follow under specific conditions (state them)
Foreclosed: This option/belief is NO LONGER available

Bedrock labels (the ONLY valid stopping points):

BEDROCK-TEST: [specific experiment or measurement]
BEDROCK-LOGIC: [logical/mathematical necessity]
BEDROCK-OBSERVE: [directly observable fact]
BEDROCK-TENSION: [contradicts established claim R-number]

Step 3: Find the Pattern

After ALL recursion is complete, look at the shape:

Expansive: The claim opens up more and more possibilities — generative, high-leverage
Constraining: The claim narrows options more and more — strong commitment, limited flexibility
Contradictory: Following implications leads to a contradiction — the claim CANNOT be fully right
Convergent: Multiple branches require the same thing — that’s foundational

Phase 2: CLAIM REGISTRY

After exploration is complete, compile EVERY claim found into a numbered list.

CLAIM REGISTRY
==============

[R1] [claim text] — TYPE: [implication/foreclosure/commitment/cost] — STRENGTH: [necessary/probable/possible]
[R2] [claim text] — TYPE: [...] — STRENGTH: [...]
...
[RN] [claim text] — TYPE: [...] — STRENGTH: [...]

Type values:

Implication: Something that must/probably/possibly follows
Foreclosure: Something that is no longer available
Commitment: Something you must also believe or do
Cost: Something you lose or must give up
Tension: Contradicts another claim in the registry

Rules for the registry:

Every R-numbered claim from Phase 1 appears here. No exceptions.
Type and strength derived from the tree, not asserted.
If a claim’s strength is unclear, mark as Possible, not Necessary.

Phase 3: SYNTHESIS

Derived entirely from the registry. No new claims introduced here.

ORIGINAL CLAIM: [X]

IMPLICATION PATTERN: [expansive / constraining / contradictory / convergent]

COMMITMENT CHAIN:
If X is right:
→ you must accept: [R-numbers of Necessary implications]
→ you probably accept: [R-numbers of Probable implications]
→ you can no longer: [R-numbers of Foreclosures]
→ you lose: [R-numbers of Costs]

WHAT THE RIGHTNESS ANALYSIS ACTUALLY FOUND:
[Numbered list of every substantive finding, referencing R-numbers]
1. [finding, from R2→R4]
2. [finding, from R6→R7]
3. ...

TENSIONS / CONTRADICTIONS:
[Any R-numbers that contradict each other. If none found, say so.]

WEAKEST LINKS:
[Which implications in the chain are Possible rather than Necessary?
 These are where the chain might break. Reference R-numbers.]

TESTABLE PREDICTIONS:
- [prediction derived from specific R-numbers]
- [prediction derived from specific R-numbers]

UNRESOLVED:
- [implications that stayed Possible — what would confirm or deny them]

Depth Scaling

Depth	Min Tree Levels	Min Implications per Node	Min Total Claims
1x	3	2	8
2x	4	2	15
4x	5	3	25
8x	7	3	45
16x	9	4	75
32x	11	4	120

Default: 2x. These are floors.

Anti-Cheerleading Checks

Failure Mode	Signal	Fix
Validation parade	Every implication is positive	Find the foreclosures and costs. What do you LOSE?
Shallow rightness	”It works, and that’s good”	Push deeper. WHY does it work? What does that commit you to?
Missing foreclosures	Only listing what opens up	Every “yes” is also a “no.” Find what closes.
Opinion bedrock	Labeling “this is good” as bedrock	Not bedrock. Keep recursing until testable/logical/observable.
Cherry-picked synthesis	Synthesis mentions 3 findings but tree had 15	Registry must include ALL claims. Synthesis references them.
Narrative tree	Tree reads as paragraphs with indent	Use numbered claims. Every node gets an R-number.

`ar - Assume Right Search`