Workflow: Run Deep Research

Use Deep Research when the question requires coordinated specialist analysis rather than a short summary.

1. Define the question

A good research question is specific about:

• subject;
• time horizon;
• decision context;
• evidence domains;
• uncertainty to resolve.

Example:

Compare the evidence supporting and contradicting the current demand narrative for NVDA over the next several weeks.

2. Choose the mode

Choose NataPulse when the answer should be grounded in curated product evidence.

Choose Normal when the investigation requires broader live web research beyond the current NataPulse dataset.

Do not describe a Normal-mode result as based only on NataPulse data.

3. Choose the scope

Select ticker, crypto, event, narrative, cluster, watchlist, market, or comparison. Enter a canonical and unambiguous reference.

4. Choose depth

• Quick: rapid structured investigation.
• Standard: balanced specialist coverage.
• Deep: broader evidence and more extensive synthesis.

Deeper does not guarantee better when evidence is sparse.

5. Choose horizon

Intraday, days, weeks, or months changes which evidence and risks matter. A short-term anomaly and a long-term fundamental thesis should not be evaluated with the same frame.

6. Add source focus and notes

In grounded mode, request on-chain, social, SEC, or macro focus when relevant. Add a precise question, constraints, or evidence that must be challenged.

7. Follow the run

Monitor phase progress and live findings. A skipped specialist can be appropriate when that domain is irrelevant or unavailable.

8. Evaluate the result

Read in this order:

1. sources and evidence quality;
2. analyst findings;
3. bull and bear debate;
4. risk review;
5. invalidation conditions;
6. final synthesis and confidence;
7. what to monitor.

9. Preserve the monitoring plan

Link the conclusion to a watchlist, alert, or later comparison. Future runs can retrieve safe lessons from the completed research.