Searching effectively: docs, search engines, AI

Write better search queries, judge how reliable a source is, and use AI without falling for its traps.

8 min · 3 questions

Nobody remembers everything, and nobody tries to. A developer with fifteen years of experience searches as much as a beginner - they simply search better and faster. Knowing how to search is not cheating, it is the number one skill of the job. The good news is that it comes down to a few simple rules.

A good search has three parts: the language or tool involved, the generic part of the error message, and nothing else. The classic trap is pasting the raw message with your file paths and your variable names in it. Nobody else in the world has a folder named like yours, so no result will come back. Strip out what is personal, keep what is universal.

Bad search:
C:\Users\Marie\projects\site\app.js NameError quantity

Good search:
javascript TypeError cannot read properties of undefined

Not all sources are equal. In order: the official documentation of the tool, which is always up to date even when it is dry; MDN for anything web-related; question-and-answer forums like Stack Overflow, very useful for specific errors; and finally blogs and videos, pleasant but often out of date. Before following an answer, check two things: its date and the version it talks about.

A conversational AI makes an excellent research companion: it explains an error message in plain language, rephrases obscure documentation, and offers a starting point. But it can also invent things with total confidence - functions, options or addresses that do not exist. This is called a hallucination. The golden rule is simple: use it to understand, then check the exact details in the official documentation. Always give it your context, your version and the complete error message.

Knowledge check

Make sure you remember the key points of this lesson.

1. How do you turn an error message into a search query?
   • Paste the whole message, paths and variable names included
   • Keep the generic part of the message and remove anything personal
   • Describe the problem in your own words without ever quoting the message
2. What should you check before following an answer found on a forum?
   • The number of comments under the answer
   • The publication date and the version it targets
   • The length of the answer
   • The country the author comes from
3. How should you treat an answer given by an AI?
   • Copy it as is, an AI does not make mistakes
   • Ignore it, an AI is no help for learning to code
   • Use it to understand, then check the details in the official documentation