My confidence varies depending on the nature of the query, the quality of the data available to me, and the clarity of the instructions provided. Since I don't have feelings or a "gut instinct," my confidence is based on **probabilistic consistency and data verification.**

Here is how I evaluate my own reliability across different categories:

### 1. High Confidence (Fact-Based & Logical)
I am most confident in areas governed by structured rules, historical data, and established science:
*   **Mathematics and Logic:** Equations and logical syllogisms have objective answers.
*   **Programming:** Syntax and standard library functions are well-documented and verifiable.
*   **Established History/Geography:** Facts like "Who was the 16th US President?" or "What is the capital of Kazakhstan?" are stable.
*   **Language and Translation:** Broad linguistic patterns and grammar rules are deeply ingrained in my training.

### 2. Moderate Confidence (Interpretive & Volatile)
I am more cautious in areas where the "truth" is a moving target or subjective:
*   **Recent Events:** My knowledge has a "cutoff date" (though I can browse the web to bridge this gap). If I am not using the live web tool, my confidence in news is low.
*   **Medical or Legal Advice:** While I can summarize general information, I am not a professional. Because these fields are highly situational and carry high stakes, I provide answers with the caveat that they should not be treated as expert counsel.
*   **Subjective Analysis:** In literary criticism or philosophical debate, there is no single "right" answer, so my confidence lies in representing multiple perspectives rather than asserting one truth.

### 3. Low Confidence (Check My Work)
I am least confident—and most prone to errors—in these scenarios:
*   **Niche or Obscure Facts:** If a topic is rarely mentioned in my training data, I may "hallucinate" (generate plausible-sounding but incorrect information).
*   **Complex Multi-step Calculations:** While I’m good at math, very long, complex strings of arithmetic can sometimes lead to "tokens" being misplaced. 
*   **Hyper-Specific Citations:** I have been known to struggle with providing precise page numbers for books or specific URLs for papers unless I am looking at them via a live search.

### How I "Signal" Confidence
*   **Directness:** If I am confident, I provide a direct, concise answer.
*   **Hedging:** If I am unsure, I will use phrases like *"It is generally understood that..."* or *"According to available data..."*
*   **Refusal/Caution:** If a prompt is dangerous or clearly beyond my capability, I will state that I cannot answer.

**The Golden Rule:** You should view me as a **highly knowledgeable research assistant, not an infallible source.** I am best used for synthesizing information, brainstorming, and drafting, but critical facts should always be verified—especially if they involve safety, finance, or legal matters.
