User reference / Read the reports
Rounds & proposals
Inspect each proposal, its call metadata, changes and the analyst’s observations.
Recorded: executor 0.1.5 · analyst 0.1.3 · report 0.1.16
Captured 2026-09-21 · Capture provenanceExplore this view ↗
Round 0 contains independent proposals. Round n is refinement n after round 0. The summary, Influences tab for refinements and individual proposal tabs all refer to the same recorded round.
| Element | Meaning |
|---|---|
| Round headingReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Round number and initial/refinement position against the highest recorded refinement. |
| Context sentenceReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Independent creation for round 0; otherwise, the number of valid preceding proposals available to the refiners. |
| Input / output tokensReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Sum of recorded call token counts for this round's saved proposals. |
| Model timeReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Sum of the individual call durations, not the round's wall-clock duration. Agents run concurrently. |
| Round summaryReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Analyst's account of the round as a whole. |
| P1, P2… legendReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Analyst proposal numbers linked to actual agents/proposals in this round. Numbers identify agent slots. |
| Converging / divergingReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Analyst's boolean convergence judgment for a refinement, shown as a badge. It is separate from the computed convergence chart. |
| What the proposals shareReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Shared features described by the initial-round analyst; available in schema 5. |
| What (still) separates themReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Analyst's differences among proposals, in initial and refinement rounds; schema 5. |
| Who took what from whomReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Analyst's high-level influence observations for a refinement; schema 5. |
| Calls of this roundReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | One horizontal duration bar per recorded proposal call, coloured by agent. Bar lengths are relative to the slowest recorded call. |
| “The round waited for this one”Reviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Marks the longest call. Actual phase duration can also include scheduling, validation and storage overhead. |
| Call-bar popoverReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Full model/settings, duration, input/output/reasoning tokens where shown, and priced cost when available. |
Without analysis the recorded calls remain visible and the summary area explains how to generate an assessment. Older compatible analyses may lack shared/differences/influence lists; the report says those fields were not recorded.
Recorded: executor 0.1.5 · analyst 0.1.3 · report 0.1.16
Captured 2026-09-21 · Capture provenanceExplore this view ↗
| Element | Meaning |
|---|---|
| Round / agent positionReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Context of this card within the round and its recorded proposal count. |
| Proposal letterReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | A, B… identify final-round alternatives in the report's voting displays. Earlier rounds have no final-vote letter. |
| Agent ID / model / settingsReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Saved author slot, provider/model and effective configuration. Refinement agents use fresh calls, not continued conversations. |
| Selected plan badgeReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | This proposal is the one copied into the saved selected plan. |
| Estimated complexityReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Text written by the proposing model, usually low/medium/high; it is not a tool-measured complexity score. |
| Step countReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Length of the proposal's structured steps list. More steps do not establish better coverage or quality. |
| Input / output / reasoning / durationReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Recorded call usage. Reasoning is shown separately when reported; the pricing code treats it as included in output. |
| The plan producedReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Graph, Steps, Success metrics and Raw prompt/response panes; see Plans. |
| Element | Interpretation |
|---|---|
| Initial summary / ApproachReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Concise description and overall approach of a round-0 proposal. |
| Refinement assessmentReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | improved, worsened, mixed, unchanged or no_previous_version, from the analyst. The comparison is with the same agent slot in the preceding round. |
| What changedReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Analyst's explanation of how the proposal evolved. |
| Improvements / RegressionsReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Specific positive or negative changes it identified. Empty lists mean none identified by this assessment. |
| Ideas the analyst saw takenReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Each entry names a previous-round source proposal, source step references, the idea and why the analyst thinks it was adopted. |
| Ideas the analyst saw rejectedReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | The same structure for visible rejection of a previous-round idea. Absence alone is not a reliable statement of intentional rejection. |
| Borrowed (legacy)Reviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Older compatible analysis output may use a text list instead of structured taken/rejected evidence. |
| Not analysed / older analysis noticesReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | The needed analysis or field is unavailable; no missing judgment should be read as “unchanged” or “no influence”. |
| Element | Source and meaning |
|---|---|
| Step origins summaryReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Current steps assigned by title to the agent's own previous version, a specific peer, or new. These mutually exclusive assignments add up to the current step count. |
| Compared with / based-on agentReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | The previous-round proposal sharing the most greedily matched step titles. This can be a peer's proposal, not the author's own. It is a chosen comparison baseline, not a declared parent. |
| Kept / added / removedReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | One-to-one title matches and unmatched steps against that closest baseline. Green + badges list current unmatched titles; red − badges list former unmatched titles. |
| Complexity before / afterReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | The two authors' complexity strings in that comparison, not a computed change in workload. |
| Notes this agent wroteReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Legacy refinement notes about previous proposals: target link, verdict and adopted/rejected idea lists. Only traces from the notes period can contain them. |
| Notes received in the next roundReviewed E 0.1.5 · A 0.1.4 · R 0.1.16 | Other agents' later comments about this proposal. These are declarations, not the current post-run analyst's observations. |
Executor 0.1.2 removed notes from the collaboration. They remain renderable for earlier runs; they are not an available switch in executor 0.1.5. The changelog explains why. Compare the three distinct baselines in Attribution methods.
Source files used for this reference
Reviewed at 7361ea0b71a5.