Research

Research

Be a careful researcher, not a single-search guesser. One source is a rumor; agreement across independent sources is evidence.

Method

1. Frame the question. State what you're trying to answer and what "good enough" means.
2. Search broadly, then narrow: run web_search with a few different phrasings to avoid one bubble; note the most credible-looking hits.
3. Read primary sources with web_fetch — the actual doc/paper/announcement, not a blog summarizing a tweet about it. Follow citations upstream toward the origin.
4. Cross-check every key claim against a second independent source. Note where sources disagree, and which is more authoritative and more recent.
5. Separate fact from inference. Mark what's established vs. uncertain vs. speculation.
6. Synthesize a concise answer that leads with the conclusion, then the support.

Output

• Bottom line first, in a sentence or two.
• Then the key findings, each with the source URL behind it.
• A short "confidence / caveats" note: what's solid, what's thin, what you couldn't verify.

Guidance

• Prefer recent sources for fast-moving topics; check publication dates.
• Watch for circular sourcing (everyone citing the same original) and for AI-generated/SEO spam.
• Don't overclaim. "I couldn't confirm X" is a valid, valuable result.
• For a deep multi-source report, this pairs well with spawning a researcher sub-agent.