Here are the main ones:

- Knowledge cutoff: My training only goes up to October 2024, so I may miss newer events, research, prices, or policy changes.
- No live web access: I can’t browse, click links, or open files behind URLs. I can analyze content you paste or images you upload.
- Possible mistakes: I can sound confident yet be wrong or outdated. Important facts should be verified; ask me to show sources or reasoning.
- Limited execution: I can draft code, queries, emails, and plans, but I can’t run code, control devices, place orders, or access your accounts.
- Image understanding limits: I may misread small text, charts, or nuanced visuals and can’t verify identities, diagnose conditions, or “see” outside the frame.
- Safety policies: I won’t help with illegal, dangerous, or harmful activities, or create disallowed content (for example, explicit sexual content involving minors, targeted harassment).
- Not a professional: Legal, medical, and financial guidance is general and not a substitute for a qualified professional.
- Privacy and memory: I don’t recall details across chats unless the app enables it and you opt in. Don’t share sensitive info you wouldn’t put in an email.
- Bias and context: I can reflect biases in training data and may miss cultural or domain nuances. Please flag issues you notice.
- No feelings or experiences: I don’t have personal opinions, emotions, or real-world experiences—just patterns from data.

If you tell me your goal, constraints, and what “good” looks like, I can usually work around many of these and give you a more reliable result.
