research（研究）

Research - Deep Investigation

Thorough research combining multiple sources and AI synthesis.

Prerequisites

• gh CLI for GitHub searches
• gemini CLI for AI synthesis

brew install gh
pip install google-generativeai
gh auth login
export GEMINI_API_KEY=your_api_key

Research Operations

Deep Research

Comprehensive multi-angle investigation:

# Step 1: Gather information from multiple angles
TOPIC="your research topic"
OBJECTIVE="what you're trying to learn"

# Web search synthesis
gemini -m pro -o text -e "" "Research '$TOPIC' comprehensively.

OBJECTIVE: $OBJECTIVE

Provide:
1. Executive summary (3-5 bullet points)
2. Key findings with sources
3. Different perspectives and debates
4. Knowledge gaps and uncertainties
5. Actionable recommendations"

Quick Research

Fast answers for straightforward questions:

gemini -m pro -o text -e "" "Quick research: [question]

Provide:
- Direct answer
- Key facts
- Important caveats
- Actionable next step"

Technology Comparison

gemini -m pro -o text -e "" "Compare these technologies for [use case]:

CANDIDATES:
1. [Tech A]
2. [Tech B]
3. [Tech C]

CRITERIA:
- Performance
- Learning curve
- Community/ecosystem
- Maintenance burden
- Production readiness

Include real-world adoption examples and common pitfalls."

Library Investigation

# GitHub search for the library
gh search repos "[library name]" --json fullName,description,stargazersCount,updatedAt --limit 5

# Check issues
gh search issues "[library] bug" --json title,url,state --limit 10

# AI analysis
gemini -m pro -o text -e "" "Evaluate [library] for production use:

1. Maturity and stability
2. Maintenance status
3. Common issues
4. Alternatives to consider
5. Recommendation"

GitHub Issues Search

Find related issues for debugging:

# Search GitHub issues
gh search issues "[error message]" --json repository,title,url,state --limit 20

# Search in specific repo
gh search issues "[error]" --repo owner/repo --json title,url,state

# Search with labels
gh search issues "[topic]" --label "bug" --state open

Research Patterns

Before Adopting a Library

#!/bin/bash
LIB="$1"

echo "=== GitHub Presence ==="
gh search repos "$LIB" --json fullName,stargazersCount,updatedAt --limit 3

echo ""
echo "=== Open Issues ==="
gh search issues "$LIB bug" --state open --json title,url --limit 5

echo ""
echo "=== AI Evaluation ==="
gemini -m pro -o text -e "" "Should I use $LIB in production? Consider maintenance, alternatives, and common issues."

Debugging with External Search

ERROR="your error message"

# Search for similar issues
gh search issues "$ERROR" --json repository,title,url,state --limit 10

# Search discussions
gh search issues "$ERROR" type:discussion --limit 5

# AI analysis of error
gemini -m pro -o text -e "" "Explain this error and likely causes: $ERROR"

Architectural Research

gemini -m pro -o text -e "" "Research best practices for: [architecture topic]

Specifically:
1. What do industry leaders (FAANG, etc.) do?
2. What are the tradeoffs?
3. Common mistakes to avoid
4. When to use vs. when to avoid
5. Concrete implementation guidance"

Staying Current

# Recent news on a topic
gemini -m pro -o text -e "" "What are the latest developments in [technology] as of 2024? Include version updates, new features, and ecosystem changes."

# Recent GitHub activity
gh search repos "[technology]" --created ">2024-01-01" --sort stars --json fullName,description,stargazersCount --limit 10

Output Synthesis

Standard Research Output

## Executive Summary
- Key point 1
- Key point 2
- Key point 3

## Detailed Findings
### Finding 1
[Details and evidence]

### Finding 2
[Details and evidence]

## Critical Analysis
- What's contested or uncertain
- What biases might exist
- What's missing

## Recommendations
1. Immediate action
2. Further investigation needed
3. Decisions to make

Comparison Output

## Quick Verdict
[1-2 sentence recommendation]

## Detailed Comparison
| Criterion | Option A | Option B | Option C |
|-----------|----------|----------|----------|
| Performance | ... | ... | ... |
| Ease of use | ... | ... | ... |

## Recommendation
[Detailed reasoning]

Best Practices

1. State objective clearly - What decision are you trying to make?
2. Use multiple sources - GitHub + web + AI synthesis
3. Check recency - Technology changes quickly
4. Verify claims - AI can hallucinate; cross-reference
5. Save valuable research - Store in memory for future reference
6. Follow up gaps - Research what's uncertain