Claude is an AI assistant made by Anthropic. You talk to it in normal language, in any language. It can write, summarize, explain, translate, plan, calculate, and reason through problems with you.
Two honest limits to know from day one:
- It can be confidently wrong. Treat important facts, numbers, and legal or medical claims as drafts to verify, not verdicts.
- It only knows what you tell it. Vague question in, generic answer out. The whole skill of using Claude is giving it enough context.
- For anything current, prices, news, opening hours, this year's rules, say "search the web for this". Claude can search, but its built-in knowledge has a cutoff date, so for fresh facts, make it look things up.
Claude does not look facts up in a database. It predicts the most likely next words from patterns it learned. That is why it can sound certain and still be wrong: confidence and accuracy are two different machines.
Go to claude.ai, sign up free, and test it on something you know deeply, your job, your hobby, your city. Ask three hard questions and grade the answers yourself.
Then get burned once, safely: ask Claude for five direct quotes from a book or song you know by heart, and check them against the real thing. At least one will likely be invented, delivered in a perfectly confident tone. That feeling is the lesson: confident is not the same as right.
Done when: you can name one thing Claude got impressively right and one thing it got wrong or fuzzy.