Fix X search returning 0 results on popular topics

Three bugs in _extract_core_subject():

1. Multi-word noise phrases ("what are", "how to") never matched
   because code compared individual words against multi-word strings.
   "what are people saying about DeepSeek R1" became "what are people
   saying" — losing the entire topic.

2. Missing meta words — "prompt", "techniques", "tips" weren't
   filtered (only "prompting" was). "vibe motion best prompt
   techniques" kept 4 keywords instead of 2.

3. No retry on 0 results — Reddit retries with simplified queries
   but X accepted 0 and moved on.

Fix: Two-phase extraction (strip multi-word prefixes/suffixes first,
then individual noise words), expanded noise set, max 3 words (was 4),
and automatic retry with first 2 words when Bird returns 0 results.

Before → After:
- "vibe motion best prompt techniques" → "vibe motion" (was 4 words, 0 results)
- "what are people saying about DeepSeek R1" → "deepseek r1" (was "what are people saying")
- "nano banana pro prompts for gemini" → "nano banana pro" (was 4 words)

Tested: vibe motion (12 X posts, was 0), DeepSeek R1 (12 posts),
kanye west (12 posts, no regression).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/plans/2026-02-07-fix-x-search-query-construction-plan.md

@@ -0,0 +1,147 @@
+---
+title: "fix: X search query too restrictive, returns 0 results on popular topics"
+type: fix
+date: 2026-02-07
+---
+
+# fix: X search query too restrictive, returns 0 results on popular topics
+
+## Problem
+
+`/last30days vibe motion best prompt techniques` returned **0 X posts** despite Vibe Motion being actively discussed on X (screenshots show posts from @Godid242, @KamilStanuch, @ColdStartTheory, @higgsfield_ai).
+
+Root cause: `_extract_core_subject()` in `bird_x.py` produces overly specific queries. Bird/X search uses **literal keyword AND matching** — ALL words must appear in a tweet. The function kept 4 keywords (`vibe motion prompt techniques`) when only 2 (`vibe motion`) were needed.
+
+## Three Bugs Found
+
+### Bug 1: Multi-word noise phrases never match
+
+```python
+# Current code (bird_x.py:24-38)
+noise = ['best', ..., 'what are', 'what is', 'how to', 'tips for', ...]
+words = topic.lower().split()  # splits into individual words
+result = [w for w in words if w not in noise]  # compares "what" against "what are" → no match!
+```
+
+`"what are people saying about DeepSeek R1"` → keeps `"what are people saying"` → **LOSES THE ENTIRE TOPIC**.
+
+The multi-word entries (`"what are"`, `"how to"`, `"tips for"`, `"use cases"`) are dead code. They never match because `.split()` creates individual words but the noise list has multi-word strings.
+
+### Bug 2: Missing meta/research words
+
+The noise list has `"prompting"` but not `"prompt"`, `"prompts"`, `"techniques"`, `"tips"`, `"tricks"`, `"methods"`, etc.
+
+- `"vibe motion best prompt techniques"` → `"vibe motion prompt techniques"` (4 words, should be 2)
+- `"nano banana pro prompts for gemini"` → `"nano banana pro prompts"` (4 words, should be 3)
+
+### Bug 3: No retry on 0 results
+
+Reddit has multi-stage retry: full query → simplified core → subreddit fallback. X search runs once and accepts whatever comes back, even 0 results.
+
+## Proposed Fix
+
+All changes in `scripts/lib/bird_x.py`.
+
+### Step 1: Fix `_extract_core_subject()` — strip phrases first, then words
+
+```python
+def _extract_core_subject(topic: str) -> str:
+    """Extract core subject from verbose query for X search."""
+    text = topic.lower()
+
+    # Phase 1: Strip multi-word prefixes/suffixes (order matters - longest first)
+    prefixes = ['what are the best', 'what is the best', 'what are', 'what is',
+                'how to', 'how do i', 'tips for', 'best practices for']
+    for p in prefixes:
+        if text.startswith(p):
+            text = text[len(p):].strip()
+            break
+
+    suffixes = ['best practices', 'use cases', 'prompt techniques',
+                'prompting techniques']
+    for s in suffixes:
+        if text.endswith(s):
+            text = text[:-len(s)].strip()
+            break
+
+    # Phase 2: Split and filter individual noise words
+    noise = {'best', 'top', 'practices', 'features', 'killer', 'guide',
+             'tutorial', 'recommendations', 'advice', 'prompting', 'prompt',
+             'prompts', 'techniques', 'tips', 'tricks', 'methods',
+             'strategies', 'review', 'reviews', 'uses', 'usecases',
+             'examples', 'using', 'for', 'with', 'the', 'of', 'in', 'on',
+             'about', 'latest', 'new', 'news', 'update', 'updates',
+             'good', 'great', 'awesome', 'and', 'or', 'a', 'an', 'is',
+             'are', 'was', 'were', 'people', 'saying', 'think', 'said'}
+    words = text.split()
+    result = [w for w in words if w not in noise]
+
+    return ' '.join(result[:3]) or topic  # Max 3 words (was 4)
+```
+
+**Expected results after fix:**
+
+| Input | Before | After |
+|-------|--------|-------|
+| `vibe motion best prompt techniques` | `vibe motion prompt techniques` | `vibe motion` |
+| `what are people saying about DeepSeek R1` | `what are people saying` | `deepseek r1` |
+| `nano banana pro prompts for gemini` | `nano banana pro prompts` | `nano banana pro` |
+| `open claw best uses` | `open claw uses` | `open claw` |
+| `best claude code skills` | `claude code skills` | `claude code skills` |
+| `kanye west` | `kanye west` | `kanye west` |
+
+### Step 2: Add retry with simplified query on 0 results
+
+In `search_x()`, after the initial search, if 0 items returned, retry with just the first 2 words of the core subject:
+
+```python
+def search_x(topic, from_date, to_date, depth="default"):
+    count = DEPTH_CONFIG.get(depth, DEPTH_CONFIG["default"])
+    core_topic = _extract_core_subject(topic)
+    query = f"{core_topic} since:{from_date}"
+
+    # ... existing Bird search code ...
+
+    items = parse_bird_response(response)
+
+    # Retry with fewer keywords if 0 results
+    if not items and len(core_topic.split()) > 2:
+        shorter = ' '.join(core_topic.split()[:2])
+        _log(f"0 results for '{core_topic}', retrying with '{shorter}'")
+        query = f"{shorter} since:{from_date}"
+        # ... retry Bird search ...
+        items = parse_bird_response(retry_response)
+
+    return response  # or merged response
+```
+
+### Step 3 (optional): Cross-pollinate Reddit entities into X Phase 2
+
+When X Phase 1 returns 0 results but Reddit found threads, extract brand/product names from Reddit thread titles and use them as X search fallback queries. This is lower priority — Steps 1-2 should fix most cases.
+
+## Acceptance Criteria
+
+- [x] `vibe motion best prompt techniques` returns >0 X posts (12 posts found)
+- [x] `what are people saying about DeepSeek R1` produces query containing "deepseek r1" not "what are people saying"
+- [x] No regressions on working queries (`kanye west`, `claude code skills`, `open claw`)
+- [x] Retry fires when initial query returns 0, logged to stderr
+- [x] `openai_reddit.py`'s `_extract_core_subject()` NOT changed (Reddit uses semantic search, not literal matching — the current function works fine there)
+
+## Files to Change
+
+- `scripts/lib/bird_x.py` — `_extract_core_subject()` rewrite + retry logic in `search_x()`
+- `scripts/lib/bird_x.py` — `search_handles()` benefits automatically (calls `_extract_core_subject()`)
+
+## Testing
+
+```bash
+# Mock mode (quick syntax check)
+python3 scripts/last30days.py "vibe motion best prompt techniques" --mock --emit=compact 2>&1
+
+# Live queries to verify X results
+python3 scripts/last30days.py "vibe motion best prompt techniques" --quick --emit=compact 2>&1 | grep -E "X:|posts"
+python3 scripts/last30days.py "what are people saying about DeepSeek R1" --quick --emit=compact 2>&1 | grep -E "X:|posts"
+
+# Regression check
+python3 scripts/last30days.py "kanye west" --quick --emit=compact 2>&1 | grep -E "X:|posts"
+```

scripts/lib/bird_x.py

@@ -24,18 +24,56 @@ def _log(msg: str):
 def _extract_core_subject(topic: str) -> str:
     """Extract core subject from verbose query for X search.
 
-    X search is literal keyword matching, not semantic.
+    X search is literal keyword AND matching — all words must appear.
-    Strip noise words to get searchable terms.
+    Aggressively strip question/meta/research words to keep only the
+    core product/concept name (2-3 words max).
     """
-    noise = ['best', 'top', 'how to', 'tips for', 'practices', 'features',
+    text = topic.lower().strip()
-             'killer', 'guide', 'tutorial', 'recommendations', 'advice',
+
-             'prompting', 'using', 'for', 'with', 'the', 'of', 'in', 'on',
+    # Phase 1: Strip multi-word prefixes (longest first)
-             'usecases', 'use cases', 'examples', 'what are', 'what is']
+    prefixes = [
-    words = topic.lower().split()
+        'what are the best', 'what is the best', 'what are the latest',
-    result = [w for w in words if w not in noise]
+        'what are people saying about', 'what do people think about',
-    # Rejoin compound terms that got split (e.g., "open claw" -> "openclaw")
+        'how do i use', 'how to use', 'how to',
-    joined = ' '.join(result[:4]) or topic  # Keep max 4 words
+        'what are', 'what is', 'tips for', 'best practices for',
-    return joined
+    ]
+    for p in prefixes:
+        if text.startswith(p + ' '):
+            text = text[len(p):].strip()
+            break
+
+    # Phase 2: Strip multi-word suffixes
+    suffixes = [
+        'best practices', 'use cases', 'prompt techniques',
+        'prompting techniques', 'prompting tips',
+    ]
+    for s in suffixes:
+        if text.endswith(' ' + s):
+            text = text[:-len(s)].strip()
+            break
+
+    # Phase 3: Filter individual noise words
+    _noise = {
+        # Question/filler words
+        'a', 'an', 'the', 'is', 'are', 'was', 'were', 'and', 'or',
+        'of', 'in', 'on', 'for', 'with', 'about', 'to',
+        'people', 'saying', 'think', 'said', 'lately',
+        # Research/meta descriptors
+        'best', 'top', 'good', 'great', 'awesome', 'killer',
+        'latest', 'new', 'news', 'update', 'updates',
+        'practices', 'features', 'guide', 'tutorial',
+        'recommendations', 'advice', 'review', 'reviews',
+        'usecases', 'examples', 'comparison', 'versus', 'vs',
+        # Prompting meta words
+        'prompt', 'prompts', 'prompting', 'techniques', 'tips',
+        'tricks', 'methods', 'strategies', 'approaches',
+        # Action words
+        'using', 'uses', 'use',
+    }
+    words = text.split()
+    result = [w for w in words if w not in _noise]
+
+    return ' '.join(result[:3]) or topic.lower().strip()  # Max 3 words
 
 
 def is_bird_installed() -> bool:
@@ -125,34 +163,17 @@ def get_bird_status() -> Dict[str, Any]:
     }
 
 
-def search_x(
+def _run_bird_search(query: str, count: int, timeout: int) -> Dict[str, Any]:
-    topic: str,
+    """Run a single Bird CLI search and return raw response.
-    from_date: str,
-    to_date: str,
-    depth: str = "default",
-) -> Dict[str, Any]:
-    """Search X using Bird CLI.
 
     Args:
-        topic: Search topic
+        query: Full search query string (including since: filter)
-        from_date: Start date (YYYY-MM-DD)
+        count: Number of results to request
-        to_date: End date (YYYY-MM-DD) - unused but kept for API compatibility
+        timeout: Timeout in seconds
-        depth: Research depth - "quick", "default", or "deep"
 
     Returns:
         Raw Bird JSON response or error dict.
     """
-    count = DEPTH_CONFIG.get(depth, DEPTH_CONFIG["default"])
-
-    # Extract core subject - X search is literal, not semantic
-    # "best open claw usecases" -> "open claw" (searchable keywords)
-    core_topic = _extract_core_subject(topic)
-
-    # Build query with date filter using X's search syntax
-    # Bird doesn't support --since flag, but X search accepts since:YYYY-MM-DD in query
-    query = f"{core_topic} since:{from_date}"
-
-    # Build command
     cmd = [
         "bird", "search",
         query,
@@ -160,9 +181,6 @@ def search_x(
         "--json",
     ]
 
-    # Adjust timeout based on depth
-    timeout = 30 if depth == "quick" else 45 if depth == "default" else 60
-
     try:
         result = subprocess.run(
             cmd,
@@ -175,7 +193,6 @@ def search_x(
             error = result.stderr.strip() or "Bird search failed"
             return {"error": error, "items": []}
 
-        # Parse JSON output
         output = result.stdout.strip()
         if not output:
             return {"items": []}
@@ -190,6 +207,47 @@ def search_x(
         return {"error": str(e), "items": []}
 
 
+def search_x(
+    topic: str,
+    from_date: str,
+    to_date: str,
+    depth: str = "default",
+) -> Dict[str, Any]:
+    """Search X using Bird CLI with automatic retry on 0 results.
+
+    Args:
+        topic: Search topic
+        from_date: Start date (YYYY-MM-DD)
+        to_date: End date (YYYY-MM-DD) - unused but kept for API compatibility
+        depth: Research depth - "quick", "default", or "deep"
+
+    Returns:
+        Raw Bird JSON response or error dict.
+    """
+    count = DEPTH_CONFIG.get(depth, DEPTH_CONFIG["default"])
+    timeout = 30 if depth == "quick" else 45 if depth == "default" else 60
+
+    # Extract core subject - X search is literal, not semantic
+    core_topic = _extract_core_subject(topic)
+    query = f"{core_topic} since:{from_date}"
+
+    _log(f"Searching: {query}")
+    response = _run_bird_search(query, count, timeout)
+
+    # Check if we got results
+    items = parse_bird_response(response)
+
+    # Retry with fewer keywords if 0 results and query has 3+ words
+    core_words = core_topic.split()
+    if not items and len(core_words) > 2:
+        shorter = ' '.join(core_words[:2])
+        _log(f"0 results for '{core_topic}', retrying with '{shorter}'")
+        query = f"{shorter} since:{from_date}"
+        response = _run_bird_search(query, count, timeout)
+
+    return response
+
+
 def search_handles(
     handles: List[str],
     topic: str,