Short answer: it depends on the task and how clearly it’s specified.

- No single global confidence: my reliability varies by domain and question clarity.
- High confidence: widely known facts up to my knowledge cutoff (Oct 2024), standard math, common programming patterns, and routine writing/editing tasks.
- Moderate: multi-step reasoning, long instructions, or code that I can’t execute to test; I may miss edge cases unless we specify tests.
- Lower: real-time or post–Oct 2024 developments, niche or very specialized topics, ambiguous questions, and subjective judgments.
- Sensitive domains (medical, legal, financial): I can provide general information, but I’m not a substitute for a professional; always verify.
- I try to flag uncertainty, state assumptions, and ask clarifying questions when needed.
- If you want, I can include a brief confidence label (high/medium/low) and key assumptions with each answer.
- You can boost accuracy by giving concrete context, constraints, examples, desired format, and by asking me to propose checks or tests.

For high-stakes decisions, please cross-check with trusted sources or experts.
