In general, my confidence varies by question type:

- **High confidence** for:
  - Well-established facts (e.g., basic math, physics principles, definitions, mainstream programming patterns).
  - Common practical tasks (e.g., writing typical code, explaining common tools, summarizing known topics).
  - Reasoning that can be done from first principles (logic puzzles, step-by-step math, structured planning).

- **Moderate confidence** for:
  - Nuanced interpretation tasks (e.g., “what does this paragraph imply?”).
  - Combining several areas of knowledge where each is well-understood, but the combination is less standard.
  - Typical “how-to” advice where multiple valid approaches exist.

- **Lower confidence** for:
  - Anything requiring **current, real-world data** after my knowledge cutoff (October 2024), like:
    - Recent news, policy changes, product releases, security vulnerabilities.
  - Highly specialized or cutting-edge research areas.
  - Questions with hidden context (e.g., legal, medical, or financial situations where small details matter a lot).

How I handle this:
- I’ll **signal uncertainty** when something is speculative, out-of-date, or depends heavily on missing context.
- I can **help you check my work**: by restating assumptions, walking through reasoning, or giving ways you could verify an answer independently.
- If you tell me the **stakes are high** (e.g., real money, health, legal risk), I’ll be more conservative and emphasize verification with a qualified human expert or up-to-date source.

If you’d like, you can give a specific type of question, and I’ll tell you how confident I’d usually be and why.
