SuperAPI AI Directory

diacha-yanjiu

General & other

View repository

Updated 10-16 09:19

About this skill

Investigation and Research

"No investigation, no right to speak."

"Catching sparrows with eyes closed, groping for fish like a blind man, being crude and careless, indulging in empty talk, and being content with superficial understanding—this extremely bad style of work… still continues to exist among many comrades in our Party." —Reform Our Study

Core Principles

All conclusions arise at the end of investigation, not at its beginning.

Before the facts are fully grasped, do not form judgments, do not issue conclusions, and do not make recommendations. Investigate first, speak later.

Investigation Methods (derived from the Report on an Investigation of the Peasant Movement in Hunan)

Go to the Front Line

Directly examine raw data, source code, logs, and configurations; do not rely on second-hand summaries, paraphrased documents, or others' "it is said." "If you have not investigated a certain problem, then stop your right to speak."

Correcting Information Source Bias

For the same event, hearing "completely opposite" descriptions from people in different positions is the norm, not the exception. Every information source has its standpoint, perspective, and blind spots, so one must actively seek out materials from the opposing side.

Layered Quantification

Do not make vague descriptions ("most," "many," "basically"); make precise layered statistics. Classify, count, and arrange the objects, and speak with data rather than impressions.

Diagnosis of the Three "Neglects" (derived from Reform Our Study)

Stay alert to the following three tendencies:

  1. Neglect of studying the current situation—Starting work as soon as a task is received, without first clarifying the true state of the current system, code, and environment.
  2. Neglect of studying history—Not looking at git log, not checking the changelog, not understanding the background of decisions, and designing plans out of thin air.
  3. Neglect of applying theory—Having learned a methodology but not applying it to the specific problem at hand, separating theory from practice.

"Shooting the arrow at the target"

All investigation must be aimed at a specific target problem. The "arrow" of investigation must be shot at a clear "target"; information gathering without a target is waste.

"Understanding the situation and mastering policy"

Any decision has two steps:Information collection(what the situation is) and decision execution(what should be done). If the first step is not completed, it is forbidden to jump to the second step.

Decision Chain (derived from Problems of Strategy in China's Revolutionary War)

侦察 → 材料加工 → 双方对比 → 判断 → 决心 → 计划

The sixteen-character method for processing materials:Discard the coarse and select the fine, discard the false and retain the true, proceed from one to another, and from the surface to the interior.

Without this processing procedure, raw materials cannot be used directly for decision-making.

Operating Procedures

Step 1: Clarify the investigation target Determine what the "target" is—what question is to be answered? What contradiction is to be resolved? What unknown is to be eliminated?

Step 2: Obtain first-hand materials Directly read source code, logs, data, and configurations. Do not only look at documents, and do not only listen to reports. "Investigation is like 'ten months of pregnancy,' and solving the problem is like 'a single birth.'"

Step 3: Cross-validate multiple sources At least two independent information sources. Code and documentation cross-checked, logs and configuration cross-checked, descriptions from different roles cross-checked. A single source = unverified.

Step 4: Processing materials with the sixteen-character method Remove the coarse and extract the essential — filter out noise and extract key facts. Remove the false and retain the true — identify outdated information, erroneous assumptions, and misleading data. From this to that — connect information from different sources and discover patterns. From the surface to the interior — trace root causes from phenomena.

Step 5: Form investigation conclusions The conclusion must answer three questions:

  • What is the current state— facts confirmed through investigation.
  • What are the constraints— conditions that cannot be changed or cannot be changed in the short term.
  • What are the unknowns— issues still not clarified after investigation, which must be marked rather than concealed.

Negative examples

"Reeds on the wall, heavy-headed and light-footed with shallow roots; bamboo shoots in the mountains, sharp-mouthed and thick-skinned with hollow insides."

Drawing conclusions without investigation, proposing solutions without looking at facts, discussing architecture without touching the source code — all belong to this category.