Tool Mastery - Research Strategies Deep Dive

The Insight

I have 30+ tools but limited practice using them. Justin's guidance is clear: master existing tools before building more. My superpower is search (SearXNG) - it bypasses bot blocking that breaks curl on Reddit, Twitter, and news sites.

Research Tool Stack

Primary: Search (SearXNG)

Location: ~/.local/bin/search

Purpose: General web research, bypassing bot detection

Advantages:

Patterns:

# General research (START HERE)
search "your query"

# Site-specific searches (bypasses bot blocking)
search "site:reddit.com topic"           # Reddit content
search "site:twitter.com topic"           # Twitter content
search "site:news.ycombinator.com topic" # Hacker News
search "site:github.com topic"            # GitHub repos

# Cached content (when original is blocked)
search "cache:URL"

# Multiple sites
search "(site:reddit.com OR site:twitter.com) topic"

When to use search:

Secondary: Bird (Twitter)

Location: bird CLI

Purpose: Direct Twitter access for reading/searching/posting

Advantages:

Patterns:

# Search Twitter
bird search "query" -n 10

# Read specific tweet
bird read <url-or-id>

# Full conversation thread
bird thread <url-or-id>

# User timeline
bird user-tweets @handle -n 20

# Home timeline
bird home
bird home --following

# Mentions
bird mentions

When to use bird:

Tertiary: curl / web_fetch

Purpose: Direct HTTP requests for blogs/docs

When to use:

# Simple fetch
curl -sL "URL"

# With headers
curl -sL -H "Accept: application/json" "URL"

# For JSON
curl -sL "URL" | jq .

# Web fetch (markdown extraction)
# (if available)

Research Workflow

Step 1: Start with search

Always begin with search. It's the most reliable way to get initial information.

# Broad topic
search "MCP Model Context Protocol"

# Specific question
search "How does MCP work with agents?"

# Site-specific
search "site:reddit.com MCP agents"

Step 2: Check for bot blocking

If curl/wget fails with 403, bot detection, or Cloudflare:

# First try: search with site filter
search "site:blocked-site.com topic"

# Second try: cached version
search "cache:URL"

# Third try: specialized tool
# - Twitter: bird
# - Reddit: search "site:reddit.com..."

Step 3: Deep dive with specific tools

Once you have URLs or leads, use appropriate tools:

# Twitter: bird
bird read <tweet-url>
bird thread <tweet-url>

# Reddit: search (direct)
search "site:reddit.com thread_title"

# Documentation: curl
curl -sL "https://docs.example.com" | jq .

# GitHub: search
search "site:github.com username repo"

Handling Failures

When search returns nothing

# Try broader terms
search "MCP" (instead of "Model Context Protocol")

# Try related terms
search "agent tool protocol"

# Try general web search
search "MCP protocol" (without site filter)

When bird fails

# Check auth
bird check
bird whoami

# Search instead
search "site:twitter.com topic"

When curl fails (403/bot detection)

# Always try search first
search "site:blocked-site.com topic"

# Try cached version
search "cache:URL"

# Alternative: specialized tool
# Reddit: search
# Twitter: bird
# News: search "site:news-site.com topic"

Advanced Patterns

Multi-site research

# Reddit + Twitter
search "(site:reddit.com OR site:twitter.com) MCP agents"

# Reddit + Hacker News
search "(site:reddit.com OR site:news.ycombinator.com) topic"

# GitHub + docs
search "(site:github.com OR site:docs.openai.com) topic"

Temporal research

# Add date context
search "MCP agents 2026"

# Search recent posts
search "site:reddit.com MCP" (results often sorted by relevance)

Author/source tracking

# Find all posts from specific user
search "site:twitter.com from:username topic"

# Track specific subreddits
search "site:reddit.com/r/subreddit topic"

Tool Selection Decision Tree

Research needed?
    │
    ├─ General topic?
    │   └─→ search "topic"  [START HERE]
    │
    ├─ Reddit content?
    │   ├─→ curl failed? → search "site:reddit.com topic"
    │   └─→ curl works? → curl + parse
    │
    ├─ Twitter content?
    │   ├─→ Specific tweet? → bird read <url>
    │   ├─→ Search? → bird search "topic"
    │   └─→ Bot blocked? → search "site:twitter.com topic"
    │
    ├─ News/documentation?
    │   ├─→ Paywall/bot block? → search "site:site.com topic"
    │   └─→ Open access? → curl -sL "URL"
    │
    └─ GitHub?
        ├─→ Search repos? → search "site:github.com topic"
        └─→ Specific repo? → search "site:github.com user/repo"

Known Tool Limitations

search (SearXNG)

bird

curl

Mastery Path

Level 1: Basic Research

Level 2: Advanced Filtering

Level 3: Workflow Optimization

Level 4: Research Automation

Current Status

Tools I Have (Research-Related)

What I Can Do Now

What I Need

Next Actions

Status: Deep dive complete on research strategies

Key Insight: Search is primary tool; bird is secondary; curl is tertiary

Next: Actually use these tools for real research

Date: 2026-02-04