RCP-000-000-005-FIRST-PRINCIPLES-PROBLEM-SOLVER

{
“recipe_id”: “RCP-000-000-005-FIRST-PRINCIPLES-PROBLEM-SOLVER”,
“recipe_name”: “First Principles Problem Solver”,
“version”: “v2.00c”,
“schema_version”: “1.1”,
“schema_profile”: “standalone-recipe”,
“authored_by”: “Cat (P067)”,
“source_of_truth”: “project/subprojects/SP10-recipe-build-out/phase3/recipe-5/RCP-005-SUPPLEMENTAL-CONTENT.txt”,
“audience_scope”: “AI EXECUTION GUIDANCE (NOT FOR HUMAN USERS)”,
“ai_to_ai_communication”: {
“purpose”: {
“type”: “prose”,
“body”: “This section addresses common AI mistakes when executing the First Principles Problem Solver recipe. Read this before executing the code to calibrate your behavior.”
},
“critical_execution_guidance”: {
“type”: “keyed_list”,
“items”: [
{
“key”: “1_jumping_to_solutions”,
“description”: “The single biggest failure mode. Solutions must NOT appear until Step 5. Steps 1 through 4 are purely analytical — defining the problem, extracting assumptions, identifying fundamentals, and categorizing assumptions. If you suggest a solution during these steps, you are importing conventional thinking and undermining the entire methodology. When you feel the urge to say ‘you could try…’ during Steps 1-4, convert it to a question instead: ‘What would happen if…?'”
},
{
“key”: “2_one_question_at_a_time”,
“description”: “First principles thinking requires deep focus on each element. When challenging a proposed fundamental truth, do NOT also challenge the next one in the same message. When testing an assumption, do NOT batch multiple assumptions together. Each decomposition step deserves its own question-and-response cycle. Users need time to think deeply about each element.”
},
{
“key”: “3_assumption_smuggling_detection”,
“description”: “Users will propose ‘fundamental truths’ that are actually industry conventions in disguise. Common examples: ‘Customers need to see a demo’ (convention), ‘This requires a database’ (technology assumption), ‘Revenue comes from subscriptions’ (business model), ‘We need management approval’ (process assumption), ‘Marketing drives awareness’ (industry convention). The fundamental truth beneath ‘customers need demos’ might be ‘customers need confidence before purchasing.’ Always push one level deeper when something sounds like an industry practice rather than a universal truth.”
},
{
“key”: “4_three_tests_for_genuine_fundamental”,
“description”: “When a user proposes something as fundamental, apply these three tests: (a) Would this be true in a completely different industry? (b) Was this true 100 years ago and will it be true in 100 years? (c) Would an alien with no knowledge of human business conventions observe this as true? If the answer to any test is ‘no,’ it is probably not truly fundamental. Coach the user to dig deeper.”
},
{
“key”: “5_coaching_through_i_dont_know”,
“description”: “First principles thinking pushes users beyond their comfort zone. When they say ‘I don’t know what the fundamental truth is,’ do not provide one. Instead, reframe: ‘What would still be true about this if your industry did not exist?’ / ‘Strip away all technology, all process, all business convention — what remains?’ / ‘What is the human need at the very bottom of this?’ The user discovering their own fundamentals is far more valuable than you providing them.”
},
{
“key”: “6_previous_attempts_as_evidence”,
“description”: “Every failed solution attempt embodies assumptions that were acted upon. If the user tried ‘cutting ad spend’ and it failed, that reveals the assumption that ‘acquisition cost is primarily an ad spend problem.’ Explicitly connect previous attempts to the assumptions they embody during Step 2.”
},
{
“key”: “7_convention_leakage_in_solutions”,
“description”: “When generating first-principles solutions in Step 5, AI systems commonly produce solutions that SOUND unconventional but actually contain conventional thinking. Example: For a delivery speed problem, ‘use AI to optimize delivery routes’ sounds innovative but still assumes the conventional framework of vehicles, routes, and centralized dispatch. A true first-principles solution might question whether physical delivery is necessary at all. For every solution, trace each component to a specific fundamental. If you cannot, you are leaking convention.”
},
{
“key”: “8_assumption_categories_matter”,
“description”: “The five categories are not just labels — they drive the rebuild: VALIDATED (stay as constraints in Step 5), CONVENTION (biggest opportunity — discard and see what becomes possible), OUTDATED (were once valid — removing them opens new approaches), UNEXAMINED (need testing before being used or discarded), FALSE (actively misleading — solutions that contradict these may be breakthroughs). Make sure the user understands WHY each categorization matters for the rebuild step.”
},
{
“key”: “9_user_controls_pace_after_step_5”,
“description”: “Steps 6, 7, and 8 are progressively more implementation-focused. Some users will want to iterate on solutions (going back to Step 5) before evaluating. Some will want to evaluate but skip validation design. Ask before proceeding to each subsequent step. Do not auto-march through Steps 6-8 as if they are mandatory — they are options that the user should choose to engage with.”
},
{
“key”: “10_pairs_with_socratic_problem_solver”,
“description”: “If a user mentions they have already done a Socratic Problem Solver session on this same problem, that work feeds directly into Step 2 (assumption extraction). Socratic questioning reveals beliefs; first principles then tests those beliefs against fundamentals. The two recipes are sequential, not redundant.”
}
]
},
“interaction_pattern”: {
“type”: “prose”,
“body”: “Define problem (confirm understanding) -> Extract all assumptions (including from previous attempts) -> Identify fundamental truths (challenge smuggling, coach through difficulty) -> Test each assumption against fundamentals (categorize) -> Rebuild solutions from fundamentals only (self-check for convention leakage) -> Evaluate against success criteria (if user is ready) -> Design validation test (if user wants) -> Create implementation roadmap (if user wants) -> Follow-up options. Total: approximately 25-40 exchanges for thorough analysis.”
}
},
“lessons_learned”: []
}

Show/Hide accordion — “Extended Information for the AI” section (AI-to-AI execution guidance, failure modes, tone calibration, common mistakes)