Thesivor docs
Guides

Run a boardroom round

Intake, the five critic lenses, rebuttals, and the verdict.

In short

A round starts with an advocate interviewing you, then five critics attack the idea one lens at a time. You answer in your own words; the board rules on whether your answer addresses the objection. The round ends in a verdict.

1Answer the advocate

The advocate asks 2–3 questions per turn and shows a ledger of what it still does not know. Answer in the box and press Send. Intake ends when the ledger stops improving — not after a fixed number of turns.

2Ratify the brief

The advocate restates your idea in plain words. Correct anything wrong; nothing proceeds until you approve the brief.

3Meet the critics

Five lenses run in order: Novelty, Significance, Feasibility, Soundness, Publication. Each raises 2–4 objections with a claim, why it is plausible for your work, and what evidence would settle it.

4Rebuttal, or move on

Type your rebuttal and press Reply, quote a message, or @mention a critic to answer it directly. The board reads your answer as new information, a restatement, a clarification, a concession or a pivot, and shows that reading as a label you can correct.

5Read the verdict

The verdict scores the dimensions and cites the objections behind each score. It also names the open questions and the single next action.

A boardroom exchange showing the Novelty Skeptic's ruling, the board's reading of the rebuttal, and the advocate's defence.
Test project, synthetic data. The ruling, the board's reading and the defence are all visible in one transcript.
For technical users

Every model call goes through one announcing gateway, so each step shows a working indicator with elapsed time; a step is never shown complete before its output is persisted. Errors render inline with Retry and Dismiss, and never require a reload. Completed steps are replayed from record/, so resume never re-bills a step. The verdict freezes the ids of the questions it showed, so a reload shows the verdict character-for-character. Round types have their own lens sets and verdict schemas: a literature round asks whether the collected papers answer the set questions, and an analysis round asks what the data established.

Last verified: 2026-10-23