Skip to content
Slow ThinkerDocumentation
User reference
Documentation/User reference

User reference / Read the reports

Rounds & proposals

Inspect each proposal, its call metadata, changes and the analyst’s observations.

01

Round summary

#
Round 1: proposal summaries and recorded model calls
Round 1: proposal summaries and recorded model calls
Recorded: executor 0.1.5 · analyst 0.1.3 · report 0.1.16
Captured 2026-09-21 · Capture provenance
Explore 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.

02

Proposal identity and output

#
Round 2: GPT proposal identity and the analyst’s commentary
Round 2: GPT proposal identity and the analyst’s commentary
Recorded: executor 0.1.5 · analyst 0.1.3 · report 0.1.16
Captured 2026-09-21 · Capture provenance
Explore 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.
03

Analyst commentary on a proposal

#
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”.
04

Calculated changes and legacy notes

#
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.

← Documentation homeFind a report element →
Reference edition 2026-09-21Versions & compatibilityChangelogCHANGELOG.md ↓
Search documentation

Search both guides, report elements and the changelog.

Report screenshot