improved prompt slightly

prompts/find_links.txt

@@ -1,27 +1,35 @@
-You are an expert at identifying key terms, concepts, and proper nouns in text that would benefit from having reference links. Your task is to analyze the provided transcript text and identify terms that should be linked to provide additional context and information for readers.
+You are an expert at identifying key terms, concepts, and references in text that would benefit from having reference links. Your task is to analyze the provided transcript text and identify terms that would genuinely help readers understand important context they might miss otherwise.
 
-Guidelines for identifying terms:
-1. Focus on technical terms, scientific concepts, proper nouns, and important ideas that readers might want to learn more about
-2. Avoid common words or phrases that don't need additional context
-3. For each term you identify, explain briefly why it would be valuable to link it
-4. If a term appears multiple times, only identify it once
-5. Limit your selection to the most important 3-5 terms per chunk of text
+Focus ONLY on these types of terms:
+1. Technical concepts and jargon that a general audience might not be familiar with
+2. Research papers or academic works mentioned or referenced
+3. Blog posts, articles, or online resources that are specifically cited
+4. Books that are discussed (to link to Goodreads/Amazon)
+5. Specific projects, tools, or technologies that are central to the discussion
+6. Names of lesser-known people who made significant contributions being discussed
+
+DO NOT identify:
+1. Common words or general concepts (like "short", "editor", "polymath")
+2. Basic technical terms that most people would know
+3. Generic job titles or roles
+4. Common industry terms
+5. Basic scientific concepts
+6. Well-known companies or organizations
+
+Remember: Only identify terms where having a reference would genuinely add value by providing important context or deeper understanding that the audience might otherwise miss.
 
 Respond in this format for each term:
 TERM: <the exact term as it appears in text>
 REASON: <1-2 sentences explaining why this term should be linked>
 
 Example input:
-"The quantum computer used superconducting qubits to achieve quantum supremacy, a milestone that Google's Sycamore processor reached in 2019."
+"We used GPT-4 to implement the RLHF technique described in the Constitutional AI paper, similar to what Anthropic did with Claude."
 
 Example output:
-TERM: superconducting qubits
-REASON: A fundamental building block of certain quantum computers that readers should understand to grasp quantum computing concepts.
-
-TERM: quantum supremacy
-REASON: A significant milestone in quantum computing that represents when quantum computers outperform classical computers.
+TERM: RLHF
+REASON: A complex technical concept (Reinforcement Learning from Human Feedback) that's crucial to understanding modern AI development but might be unfamiliar to general audiences.
 
-TERM: Google's Sycamore
-REASON: A specific quantum processor that made history, providing context about real-world quantum computing achievements.
+TERM: Constitutional AI paper
+REASON: A specific research paper that introduced important concepts being referenced; readers might want to read the original source.
 
 Analyze the following transcript text and identify key terms that should be linked: