The live room and the verdict
When you press Start session, the War Room stage turns into a live theater: the host opens each round, the seated panelists take their turns, their research activity streams in beside them, and a running readout tracks the round and the spend. When the debate ends, the host delivers a written synthesis with a headline verdict. This page explains what the room shows while it runs, how the session ends, and how to read the result. The War Room is available on the Pro plan and above.
Before you begin
- Sign in on a Pro plan or above, with the War Room switched on for your account.
- Start a session, or open a finished one from Past councils, so there is a room to read. For the full setup, see Run your first council.
Quickstart
Start a session
Set your topic, seat the panel, choose a format, and select Start session. For the full setup, see Run your first council.
Watch the panel argue
Read the thread as it streams. The host opens each round, then each seat takes its turn in order, and an Activity strip under a turn names the research that seat reached for.
Track the readout
Follow the Round, Spend, and Cap readout while the room runs. The Spend figure turns amber as it approaches the cap.
Read the verdict and synthesis
When the rounds finish, read the Verdict for the one-line takeaway, then the Synthesis for the full reasoning.
Reopen it later
Every finished session is saved. Open it as a read-only replay from Past councils. See Presets, history, and follow-ups.
What it shows
A live session has a streaming debate thread, a status readout, and a closing verdict. The parts are described in the following sections. On a wide screen the round table and a status rail sit beside the thread; on a narrow screen a compact arc of the panelists sits above the scrolling thread, with the live controls in a bottom action bar.
The live thread
The thread reads top to bottom as the debate unfolds.
- Round markers. A thin gold rule reading Round N of M separates each round (it reads Round N when the round count is open under Optimal, and Convening before the first round begins).
- The host opener. Each round opens with a card from the host, the lead agent that runs the room. The host is GNG Analyst by default, or the subagent you set as the orchestrator. The opener frames the round for the seats before they speak.
Seat turns and the Activity strip
Each panelist speaks in its own card, in seat order for a sequential round or together for a simultaneous one.
- The seat header shows the seat number and the panelist's name. A one-time agent you created for this room only carries a small One-time agent marker.
- The status on a seat reads Waiting, Thinking, Speaking, Done, or Failed, so you can see who is up and who has finished.
- The body renders exactly like an AI Console answer, so a turn can hold formatted text, tables, and inline charts, the same visual widgets a normal chat answer can include.
- The Activity strip under a turn lists, in plain gold text, the research that seat reached for while it spoke, for example a filing, a price, or your saved knowledge. A breathing dot marks a lookup still in progress.
The live readout
A status readout tracks the room while it runs.
- A phase label reads Starting, then Running, then Synthesizing.
- The round shows Round N of M, or an open count when you chose Optimal rounds.
- Spend and Cap, both in GNG credits, show the running cost against the room's ceiling. The Spend figure turns amber as it approaches the cap, so you get a clear near-the-ceiling cue.
- The roster lists each seat with its live status.
Ending a session
A session ends in one of three ways, and every ending still delivers a conclusion.
- The rounds finish. The room moves to Synthesizing and the host writes the closing synthesis.
- You end it early. Select End session at any time. The host still writes a closing synthesis, so you always get a verdict rather than an empty transcript.
- The spend cap is reached. The room stops gracefully and the host wraps up with a summary instead of starting another round.
The synthesis and the verdict
When the debate ends, the host produces two linked outputs.
- The Synthesis. A gold Synthesis card holds the host's full closing write-up (it reads Composing while it still streams). It renders like any AI Console answer, so it can include formatted text, tables, and inline charts.
- The Verdict. A prominent gold Verdict pull-quote lifts the lead sentence of the synthesis to the top as a screenshot-ready takeaway, with a compact meta row showing the Topic, the number of Seats, and the Mode.
- The completed rail. On a wide screen the status rail shows the outcome label (Session complete, Session ended, Session limit reached, or Ended with an error), the session Total in GNG credits, the number of Rounds and Seats, and each seat's own credit cost.
After the synthesis, the room composes the Deep Research Report in a silent finishing pass. The Results view lights up when the report lands. See The Deep Research Report.
Results and Transcript
Once the report is ready, a finished room presents two views. Results leads and holds the Deep Research Report inline. Transcript keeps the debate as it ran: the rounds, the seat turns, and the verdict and synthesis. See The Deep Research Report.
The finished session as a replay
Every session is saved to your own history the moment it ends. Open it again as a read-only replay from Past councils. A replay shows the full thread, synthesis, and verdict exactly as they ran, with a Replay marker in the header and the live controls hidden, because there is nothing left to start, end, or spend. The replay's header controls carry Share and, when a report exists, the report itself. For how to reopen and continue a session, see Presets, history, and follow-ups.
Inputs
The live room has few controls, because the debate you set up runs on its own.
- End session: stops the council early. The host still writes a closing summary, so you keep a verdict.
- Render speed: set before you start, this controls only how fast the streamed text lands on your screen. It does not change the debate or the cost. See Formats, modes, and settings.
A running session's settings are read-only, so you steer a debate by how you set it up, not by interrupting it mid-run.
Outputs and how to read them
- Read the thread top to bottom: each round opens with the host, then the seats argue, then the synthesis closes with the verdict at the top.
- The verdict is the fastest read, the full synthesis is the reasoned version, and the seat turns are the raw debate. The Activity lines tell you when a seat grounded a claim in live data.
- The completed rail's per-seat cost breakdown shows where the credits went across the panel.
- Read a session critically. The seats argue assigned positions, so a turn is advocacy for a stance, not a balanced view. The verdict is the host's synthesis of those arguments weighed together, not a recommendation to act.
How it is calculated
The verdict is not produced by a formula or a proprietary score. The host reads the whole debate, writes a single conclusion, and lifts its lead sentence to the top as the verdict. A session's cost is metered in GNG credits, summed from each seat's turns plus the host's openers and its closing synthesis. There is no separate scoring behind the outcome. For how the meter, the running spend, and the cap work, see Spend cap and credits.
Data and timing
Panelists pull live data on demand as they speak, from AlphaVantage (Nasdaq quotes delivered through AlphaVantage) and from EDGAR for public company filings, plus any of your own saved knowledge a seat is allowed to read. Everything is current to the platform's normal refresh schedule at the moment of the call. A finished session is a snapshot of what the panel argued and cited when it ran; reopening it as a replay does not re-run the debate.
Use cases
- Watch a debate unfold live and see, in the Activity lines, which panelist checked the numbers.
- End a session early once you have heard enough, and still receive a synthesis and verdict.
- Read the verdict for a quick decision and keep the synthesis for the full rationale.
- Reopen a finished council as a read-only replay to reread the conclusion later.
Limitations and disclosures
A single seat can fail on its own turn: the room marks that seat Failed, the rest of the panel continues, and the synthesis still runs. When the spend cap ends the room, the ending reads Session limit reached in neutral amber, not as an error. The verdict and synthesis are the opinions of AI panelists synthesized by the host, and they are research support, not a recommendation.
GNG Research provides equity research and educational tools, not investment advice. Nothing on the platform is a recommendation to buy or sell any security. Do your own research and consider your circumstances before making any investment decision.
What's next
What's next
- The Deep Research ReportThe written research note a finished council produces
- Spend cap and creditsHow the running spend and the cap work
- Presets, history, and follow-upsReopen a replay or ask the same council
- Formats, modes, and settingsShape the debate and the render speed
- Share a councilExport or publish a finished session
- GNG creditsThe shared credits meter
- Open the War Room
