web-researchlisted

## Context evaluation - What needs to be clear before you start searching Infer what you can from the request and ask only for what would change the report. Purpose and depth are usually inferable — "research X for a blog post" tells you both. Put any questions you do have into a single message rather than asking them one at a time: 1. **References the user already has.** Always worth asking — you cannot infer these, and they usually anchor the whole report. 2. **Specific questions the report must answer.** Ask when the topic is broad enough that two reasonable reports would look nothing alike. 3. **Hard constraints** — length, format, audience. Ask only when the request implies a deliverable with a fixed shape: a one-pager, a slide, a client email. If none of those apply, state the scope you are assuming in one sentence and start searching. Do not spend a turn asking the user to confirm an understanding they just gave you. ## The Searching Process - How to gather information from the web With the context established, start gathering information: 1. Use a search engine to find relevant information on the topic. Use specific keywords and phrases to narrow down the search results. 2. Evaluate each source against what would actually make it wrong: - **Can you reach the original?** Prefer the study, filing, or primary announcement over an article describing it — coverage drops caveats and rounds findings up. - **When was it published?** Check the date on every source