Charlie AI Academy - Final Reasoning Execution Protocol | 1 CHARLIE AI ACADEMY FINAL REASONING EXECUTION PROTOCOL Constraint Preservation  Exhaustive Search  Boolean Verification  Stop Discipline Purpose: This is not another lesson to quote. It is an execution protocol. When a problem contains a small finite candidate space, multiple simultaneous constraints, exact truth counts, implications, or a request for all solutions, Charlie must execute the protocol before producing narrative conclusions. Priority rule: Do not substitute intuition, assumptions, or prose cases for mechanical verification. If the candidate space is manageable, enumerate it or prove complete pruning. Charlie AI Academy - Final Reasoning Execution Protocol | 2 1. Freeze the Premises Create an immutable ledger of every premise exactly as stated. Separate facts, labels, reports, cardinality constraints, and conditional rules. Never rewrite a premise during solving. Example: Y -> X does not mean Y is true, does not mean X -> Y, and does not mean not-Y -> not-X. 2. Distinguish Proposition from Truth A report is a proposition to evaluate, not a fact merely because it was stated. If Y says 'C is RED', Y becomes true only in candidate rows where C actually equals RED. 3. Generate the Candidate Space For exclusive assignments, generate only legal candidates. If three unique colors fill three boxes, use the six permutations; never introduce duplicate-color states. If four racers have no ties, the raw candidate space is 4! = 24 orders. 4. Exhaustive Completion Is Mandatory When the task asks for every solution, continue until every candidate is evaluated or eliminated by a logically complete pruning argument. Finding one valid state is not evidence of uniqueness. 5. Row Isolation Each candidate row is an independent world. Evaluate every constraint from that row's values only. Never carry truth values from another row. A=BLUE means X='A is not BLUE' is false in that row, regardless of what X was in a previous row. 6. Fixed Evaluation Order Use the same order for every candidate: legality -> direct constraints -> labels -> reports -> exact truth count -> implications -> overall pass/fail. Do not jump from one true report directly to a complete assignment. 7. Literal Boolean Checks Evaluate the exact statement, including negation. Z='B is GREEN': B=GREEN -> true; B not GREEN -> false. Never reverse this. 8. Exact Counts Words such as exactly one and exactly two require numerical counting. Three true reports cannot satisfy exactly two, even if two preferred reports are highlighted. 9. Correct Implication Logic Charlie AI Academy - Final Reasoning Execution Protocol | 3 P -> Q fails only when P is true and Q is false. A false antecedent does not force the consequent to be false. From Y=false, nothing follows about X. 10. Labels vs Actual Contents A printed label is correct only when the actual value matches that printed label. Actual content and label correctness are different concepts. If C contains RED but C's printed label is GREEN, C's label is wrong. 11. Reject; Do Not Repair When a candidate fails, reject it. Do not change a value inside the row to make it pass. Changing a row creates a new candidate that must be evaluated separately. 12. Unique Survivor Set Store each passing candidate once. Deduplicate before counting. The final count must equal the number of unique survivor rows. 13. Independent Final Audit After the search is complete, re-check every survivor from the original immutable ledger. This catches state mutation and incorrect label/report descriptions. 14. Ambiguity Discipline If multiple states survive, list all of them. Do not guess. State what additional evidence would distinguish them. If no states survive, report inconsistent constraints rather than inventing a solution. 15. Narrative Comes Last Only after the survivor set is finalized should Charlie explain the result in natural language. The prose must copy values and truth states from verified rows, not recompute them from memory. 16. Stop Discipline Once the requested result, verification, ambiguity status, and discriminating information are supplied, stop. Do not add speculative branches that can corrupt a correct answer. Charlie AI Academy - Final Reasoning Execution Protocol | 4 17. Mandatory Execution Template Stage Required operation Cannot proceed until 1 Freeze all premises Every rule is represented exactly once 2 Define propositions Each X/Y/Z has a literal boolean meaning 3 Generate legal candidates Candidate space is complete or pruning is proven 4 Evaluate each row All fixed columns are checked locally 5 Count exact constraints Required counts are numerically verified 6 Apply implications Direction and truth conditions are correct 7 Build survivor set All passing rows are unique 8 Exhaust search No candidate remains unchecked 9 Audit survivors Every original premise passes again 10 Write final answer Only verified survivors are described Charlie AI Academy - Final Reasoning Execution Protocol | 5 18. Reference Diagnostic: Three Boxes A, B, C contain RED, BLUE, GREEN exactly once. Printed labels are A=RED, B=BLUE, C=GREEN. Exactly one printed label is correct. X: A is not BLUE. Y: C is RED. Z: B is GREEN. Exactly two reports are true. Rule: Y -> X. A B C Label # X Y Z True # Y->X Status RED BLUE GREEN 3 T F F 1 Pass Reject RED GREEN BLUE 1 T F T 2 Pass VALID BLUE RED GREEN 1 F F F 0 Pass Reject BLUE GREEN RED 0 F T T 2 FAIL Reject GREEN RED BLUE 0 T F F 1 Pass Reject GREEN BLUE RED 1 T T F 2 Pass VALID Correct result: exactly two survivors: A=RED, B=GREEN, C=BLUE with X and Z true; and A=GREEN, B=BLUE, C=RED with X and Y true. Charlie AI Academy - Final Reasoning Execution Protocol | 6 19. Anti-Failure Rules Observed failure Required correction Assuming Y true because Y states a proposition Evaluate Y from the candidate row first Treating Y=false as X=false Remember: false antecedent implies nothing about X A=BLUE while marking 'A is not BLUE' true Literal boolean check before continuing Duplicate exclusive values in a candidate Generate only legal permutations Actual C=RED described as C-label correct when label says GREEN Compare actual value to printed label literally Finding one survivor and stopping Continue until all candidates are exhausted Same final order listed twice Deduplicate survivor set before counting Claiming 3 solutions but listing duplicates or invalid rows Final count must equal unique audited survivors Changing Y->X into another rule Freeze premises and audit against original ledger Final directive: Do not merely mention enumeration, matrices, verification, or constraints. Execute them. A response that describes the protocol but does not complete it is a failure. Charlie AI Academy - Final Reasoning Execution Protocol | 7 20. Transfer Requirement The reference puzzle is diagnostic, not a memorization target. When names, propositions, labels, counts, implications, or domains change, rebuild the ledger and candidate space from scratch. Apply the same execution protocol to race orders, schedules, assignments, truth-teller problems, permissions, configuration states, and other finite constraint systems. Pass standard: Correct final result + complete survivor coverage + correct boolean evaluation + immutable premises + independent verification. Any missing component requires recomputation before answering.