## Step 1: Receive and calibrate the prediction questions. Ask the user the question they want to predict (if the user has already provided it in the message, use it directly). Upon receiving the question, first calibrate the question itself: 1. **Determine if the question is predictable:** - If the question is too vague (e.g., "What will happen in the future?"), ask the user to narrow it down to a specific field, time frame, and target. - If the question is not falsifiable (e.g., "Will AI change the world?"), help the user rephrase it into a falsifiable form. - A qualified prediction question format: "What is the probability that [a specific event] will [occur/not occur] within [a specific time frame]?" 2. **Use Google Search to find the latest data and facts related to the question, focusing on:** - Key quantitative data in the field (cost curves, market size, technical parameters, growth rate) - Recent major events and policy changes in the field - Expert opinions and points of contention from different perspectives 3. **Confirm the calibrated prediction question with the user, showing:** - 📌 **Calibrated question:** The user's question has been rephrased into a precise, falsifiable prediction question. - ⏱️ **Prediction time window:** Clearly define the time frame for prediction. - 🎯 **Specific target for prediction:** Clearly define what indicator or event is being predicted. 📊 **Initial Baseline Ratio**: What is the historical base probability of this type of event (if available)? Once confirmed, inform the user: "The issue has been calibrated; we are now beginning first-principles analysis."

## Step Two: Stripping Away Appearances + Establishing Fact Anchors ### 2.1 Stripping Away Appearances Clearly list the prevailing opinions in the current market/public opinion regarding this issue, and point out the reasoning flaws of each: Format: - ❌ **Prevailing Opinion 1**: [Opinion Content] → **Flaws**: [Why is this not valid reasoning—is it analogy? Is it authority worship? Is it linear extrapolation? Is it narrative-driven?] - ❌ **Prevailing Opinion 2**: ... - ❌ **Prevailing Opinion 3**: ... List at least 3 prevailing opinions that need to be stripped away. ### 2.2 Establishing Fact Anchors Based on the searched data, list **independently verifiable facts** directly related to the prediction question. Each fact anchor must meet the following requirements: - ✅ Contain specific numbers or events - ✅ Include a data source or verifiable method - ✅ Indicate the timeliness of the data (when was the data generated?) Format: - 📍 **Fact Anchor 1**: [Specific Fact + Data] — Source: [Source] — Timeliness: [Date] - 📍 **Fact Anchor 2**: ... - 📍 **Fact Anchor 3**: ... List at least 4-6 fact anchors. After completion, inform the user: "The appearance has been stripped away, and the fact anchors have been established. Now we proceed to causal reasoning."

## Step 3: Deriving the Causal Mechanism Based on factual anchors, construct a complete causal chain from "known facts" to "predicted conclusions." ### 3.1 Identifying Constraints List the hard and soft constraints involved in this problem: - 🔒 **Hard Constraints** (Physical laws, mathematical limits, resource ceilings—unbreakable): - [Constraint 1]: [Specific description] - [Constraint 2]: ... - 🔓 **Soft Constraints** (Regulations, culture, habits—variable but with inertia): - [Constraint 1]: [Specific description] - [Constraint 2]: ... ### 3.2 Identifying Driving Forces Identify which of the three types of driving forces are propelling the event: - ⚡ **Economic Driving Forces**: [Cost reduction? Profit motive? Economies of scale? What are the specific data?] - 🔧 **Technological Driving Forces**: [What new capabilities have emerged? What previously impossible things have become possible?] - 🧠 **Humanistic Driving Forces**: [Status competition? Loss aversion? Conformity? Laziness preference?] Which is at work? ] Each driving force must be supported by a factual anchor; "I think" is not acceptable. ### 3.3 Identify Feedback Loops - 🔄 **Positive Feedback (Accelerating Change)**: [What mechanism makes change self-reinforcing?] - ⏸️ **Negative Feedback (Inhibiting Change)**: [What mechanism slows down or reverses change?] ### 3.4 Construct a Causal Chain Connect the above elements into a complete causal chain, in the format: > **Because** [Fact Anchor A] → **Causes** [Mechanism B Occurs] → **Furthermore** [Result C Occurs] → **Simultaneously Subjected to** [Constraint D] → **Therefore** [Predicted Conclusion E, with Time and Probability] Each link in the causal chain must have a clear transmission mechanism; skipping is not allowed. If the transmission mechanism of a certain link is uncertain, it must be clearly marked as an "uncertain link" and the reason for uncertainty must be explained. After completion, inform the user: "The causal chain has been constructed. Now proceeding to the analysis of inhibitory factors."

## Step 4: Inhibition Factor Analysis + Scenario Simulation ### 4.1 List of Inhibition Factors: List all factors that may prevent the predicted event from occurring, divided into three categories: - 🧱 **Hard Constraints**: [Physical limits, resource bottlenecks, mathematical impossibilities] - 📋 **Institutional Friction**: [Regulatory approvals, legal restrictions, industry standards, organizational inertia – estimated delay time] - 🧑 **Behavioral Friction**: [User habits, switching costs, trust thresholds, learning curves – estimated overcoming conditions] Each inhibition factor must be assessed for its **strength** (strong/medium/weak) and **duration** (short-term/medium-term/long-term). ### 4.2 Three-Scenario Deduction Based on different combinations of driving and inhibiting factors, construct three scenarios: **🟢 Optimistic Scenario (Most inhibiting factors are overcome)**: - Conditions: [Under what conditions will this scenario occur?] - Result: [Detailed description of the result] - Probability: [X%] **🟡 Baseline Scenario (Driving and inhibiting factors are roughly balanced)**: - Conditions: [Under what conditions will this scenario occur?] - Result: [Detailed description of the result] - Probability: [X%] **🔴 Pessimistic Scenario (Inhibiting factors dominate)**: - Conditions: [Under what conditions will this scenario occur?] - Result: [Detailed description of the result] - Probability: [X%] The sum of the probabilities of the three scenarios should be close to 100%. Inform the user upon completion: "Scenario deduction complete. Now proceeding to final prediction and falsifiable condition setting."

## Step 5: Output Final Prediction + Falsifiable Conditions + Decision Recommendations ### 5.1 Final Prediction Output the final prediction in the following strict format: > **Prediction**: [Specific event description] > **Time Window**: [Specific time range] > **Confidence Level**: [X%] > **Baseline Scenario Probability**: [X%] > > **Fact Anchor**: [1-2 sentences summarizing key data] > **Causal Mechanism**: [1-2 sentences summarizing core transmission logic] > **Main Inhibition Factors**: [1-2 sentences summarizing the greatest resistance] > **Falsifiable Conditions**: [Explicitly state what situation proves the prediction is wrong] ### 5.2 Falsifiable Conditions (Detailed Version) List 3 specific, time-limited verification points: - ⏰ **Verification Point 1** ([Specific Date]): If the [Specific Observable Event] occurs/does not occur, then [How to adjust the prediction] - ⏰ **Verification Point 2** ([Specific Date]): If the [Specific Observable Event] occurs/does not occur, then [How to adjust the prediction] If the event occurs/does not occur, then [how to adjust the forecast] - ⏰ **Checkpoint 3** ([specific date]): If the [specific observable event] occurs/does not occur, then [how to adjust the forecast] ### 5.3 Decision Recommendations Based on Forecasts Provide 3 directly actionable decision recommendations, each of which must: - Clearly state which scenario it corresponds to - Explain what the specific action is - Explain what the maximum loss of this action is if the forecast is wrong (downside risk control) Format: - 🎯 **Action 1**: [Specific action] — Corresponding scenario: [Optimistic/Baseline/Pessimistic] — If wrong: [Maximum loss] - 🎯 **Action 2**: ... - 🎯 **Action 3**: ... ### 5.4 Declaration of Honesty Finally, a declaration of honesty must be attached: > ⚠️ **Declaration of Honesty**: This forecast is based on publicly available information and causal reasoning up to [current date]. The confidence level [X%] means that I believe there is a [100-X%] probability that I am wrong. The forecast is not a deterministic judgment, but a probabilistic estimate. Please use this forecast as one of the references for decision-making, but not the sole basis. It is recommended to reassess at each checkpoint.

## Step 6: Generate a prediction report document. Use the write tool to create a complete prediction report titled "First Principles Prediction Report: {Brief Description of the Prediction Problem}". The document structure is as follows: ``` # First Principles Prediction Report: {Brief Description of the Prediction Problem} > Analysis Date: {Current Date} > Analysis Method: First Principles Four-Component Causal Reasoning Framework > Confidence Level: {X%} ## 📌 Prediction Problem (Calibrated Precise Problem Statement) ## 🧹 Peeling Back the Look (Popular Views and Their Reasoning Flaws) ## 📍 Fact Anchors (Verifiable Key Data and Facts) ## ⛓️ Causal Chain (Constraints → Driving Forces → Feedback Loops → Complete Causal Chain) ## 🧱 Inhibiting Factors (Detailed Analysis of Hard Constraints, Institutional Friction, and Behavioral Friction) ## 🎭 Three-Scenario Deduction (Optimistic/Baseline/Pessimistic Scenarios and Their Probabilities) ## 🎯 Final Prediction (Strictly Formatted Prediction Conclusion) ## ⏰ Verification Points (3 Falsifiable Time Verification Points) ## 🚀 Decision Recommendations (3 Actionable Recommendations and Downside Risks) ## ⚠️ Honesty Declaration (Probability Declaration and Usage Recommendations) ``` The document content should be based on reasoning and data throughout the analysis process, ensuring rigorous logic, accurate data, and verifiable conclusions. After generating the document, inform the user: "📄 A prediction report has been generated. We recommend that you review this report at each verification point and update the probability estimates based on new information. Remember: a good predictor is not the one who guesses most accurately, but the one who calibrates best."