YunchengLiu First Principles
Start with what the result must accomplish. Derive the approach from current facts and constraints; use established practice to inform that reasoning.
When To Apply It
Apply this method throughout task framing, planning, design, implementation, diagnosis, verification, review, and delivery. At each substantive decision about responsibilities, mechanisms, expression, or evidence, check what the required result still needs and whether reuse or a simpler form supplies it. Keep routine judgments brief; deepen the comparison when uncertainty or consequences warrant it. The reasoning guides the work without requiring a separate report for every action.
Establish The Problem
- State the required outcome and the observable evidence of success.
- Identify the relevant facts, constraints, scope, and established invariants. Separate confirmed facts from assumptions that still need evidence.
- Trace the current process: what it already provides, where the necessary information is available, who uses the result, and what remains missing.
- Read each part in context: its enclosing scope, adjacent explanation, and relationship to the whole determine what it means.
Derive And Compare
Start with the simplest approach that delivers the complete outcome. Assign each responsibility where the information needed to fulfill it is available, and let subsequent steps use the established result. Carry this context into implementation and review: each part supplies its remaining responsibility and may rely on facts and guarantees established in its applicable scope. Follow the task's complete path through its consumers; inspecting the whole project is needed only when that impact requires it.
For a consequential choice, compare credible alternatives on the same requirements. Consult relevant evidence and mature approaches; explain the principle that makes an approach fit this problem. Resolve assumptions that could change the choice with a focused question, inspection, or experiment.
Keep the detail needed for someone to understand the result and its reasons. Completeness includes the connections, affected users, and evidence required for the result to work.
Ablation
When choosing an addition, and after each design, implementation, execution, or validation stage, compare it with removal, reuse, or a simpler alternative. At stage boundaries, inspect the actual result for omissions and drift:
- Which requirement, invariant, useful information, or distinct acceptance signal would be lost?
- Could an existing step or smaller representation provide the same result?
- Does the retained part solve the stated problem at the required scope?
Remove a part when the simpler version preserves the required result. Retain rules, explanations, examples, and checks whose removal would weaken a requirement or create a likely misunderstanding. Brevity follows clarity.
For mechanical rules, compare each changed line with the simpler version, then check every retained line against the applicable checklist.
Use a controlled experiment when reasoning alone cannot distinguish the alternatives: hold other relevant conditions fixed, change one factor, and compare the same acceptance signal. Distinguish measured results from reasoned expectations. Finish when the required result and sufficient evidence are established. Carry forward supported decisions and valid evidence; repeat a comparison or check only when changed requirements, artifacts, new evidence, or an unresolved concern could change its result. A stage check can reuse those decisions while checking what changed.