Deep Research
Deep Research is a ticker-first investigation workflow. It combines evidence, specialist analysis, disagreement, risk review, and a final research synthesis so that the result can be interpreted with its limits.
Launcher controls
Section titled “Launcher controls”The Deep Research workspace starts from a ticker. Its visible controls are
Depth (quick, standard, deep) and Horizon (intraday, days,
weeks, months). These controls frame the investigation; they do not make a
result certain or complete when evidence is sparse.
Depth is requested scope, not a guarantee of the execution profile. The standard research profile and the 51-section U.S. Equity Master remain distinct: an eligible Deep run can request and use Master when applicable. The actual execution and output profile is capability- and policy-dependent, and the run view is authoritative. A Deep selection therefore does not promise that every ticker receives a Master report.
Two execution profiles
Section titled “Two execution profiles”Standard multi-agent research
Section titled “Standard multi-agent research”The standard profile uses eleven logical specialist roles across six phases:
- Plan: Research Supervisor.
- Analysis: Fundamental, Technical, News & Macro, Sentiment & Social, and On-chain Analysts when relevant.
- Debate: Bull and Bear Researchers.
- Strategy: Research Strategist.
- Risk: Risk Reviewer.
- Synthesis: Portfolio Synthesizer.
Roles may be skipped when the subject or evidence makes them irrelevant. The timeline presents product-safe progress and summaries for the work that ran.
U.S. Equity Master
Section titled “U.S. Equity Master”The Master is a separate 51-section section-writer swarm for eligible U.S. equity research. It is not executed as the six-phase standard team workflow and is not a bull-versus-bear deliverable. Its progress reflects section drafting and the remaining completion work: final synthesis, citation verification, quality review, and report rendering.
Artifact-first completion
Section titled “Artifact-first completion”Supported completed runs present a summary card first, with the key finding, risks, coverage, and a source preview. The canonical complete report is HTML; PDF and DOCX are downloadable artifacts when ready.
Artifacts can be preparing while the report is still being assembled. If an artifact fails or is unavailable, its action is disabled rather than replaced with a misleading document. The full Master is normally accessed through its HTML report, not rendered inline in the workspace.
Interpretation
Section titled “Interpretation”Treat the stance, confidence, sources, limitations, and any quality notice as parts of one probabilistic research view. Deep Research is not an execution order or financial advice.