The forty-five-minute reading companion presents the question, evidence gaps, competing explanations, and a proposed comparison in four panels. It shows Agreement and Observation as unanswered labels. A single notes box supports discussion; the actual agenda and any later decisions belong in your maintained meeting document.
Use the worked example
Select Competing explanations to compare E1, configuration contributes, with E2, warmup contributes. Select Proposed checks to read P01's matched-warmup idea. Use the notes box for questions raised during the discussion. Copy anything needed into your Markdown or DOCX agenda before leaving the page; Reset clears these scratch notes.
Keep the decision attached to its evidence
A stuck comparison is a useful meeting topic even when its explanation remains open. Cornell's lab meeting guidance includes work in progress, uncertain figures, methods, and unresolved problems. Bring the question and the approaches already tried so colleagues can help identify the next discriminating observation.
Write the question before the timetable: Does the current runtime comparison distinguish a configuration effect from differences in warmup?
For your own meeting, add the presenter, note taker, date, and pre-read reference to the maintained agenda. Keep the prepared question separate from notes written during the discussion. When no check owner has been agreed, record that gap instead of treating the proposed comparison as an assigned experiment.
Keep the 20/28 ms comparison and warmup limitation with the current pre-read reference in the maintained meeting packet. In the reading view, the Evidence panel makes missing method and raw-artifact information visible before anyone chooses a candidate explanation.
The arithmetic gap is 28 − 20 = 8 ms in the supplied fictional observations. It does not establish that changing configuration caused an eight-millisecond effect. Do not label A Faster, Recommended or Validated simply because its stated value is lower.
The explanations are candidates, not established causes or an exhaustive list. Both could contribute; other conditions may matter. A repeat must record what it holds constant and what it changes before anyone interprets an outcome.
Ask what observation would weaken each explanation and which evidence is still unavailable. That directs discussion toward distinctions a check can address, rather than a vote for the lower number.
The five agenda segments total 5 + 10 + 15 + 10 + 5 = 45 minutes. Adjust that illustrative allocation to the question your group needs to discuss. Keep the slot focused on the comparison rather than adding unrelated items to fill time.
In the final five minutes, distinguish what was discussed from what was agreed. If no protocol or owner was agreed, record that. An unresolved close with an explicit evidence gap is useful; a prepared agenda cannot supply consensus.
| Original evidence field | Configuration A | Configuration B | What it establishes |
|---|---|---|---|
| Supplied illustrative runtime | 20 ms | 28 ms | Two stated observations, not a validated benchmark |
| Configuration/method version | TO CONFIRM | TO CONFIRM | Exact conditions not supplied |
| Warmup | Reported different | Reported different | Counts not supplied; comparison condition differs |
| Input set/protocol | TO CONFIRM | TO CONFIRM | Matching inputs/procedure not established |
| Replicate count/statistic | TO CONFIRM | TO CONFIRM | Cannot call these a mean, median or reliable distribution |
| Raw artifact/version reference | Not supplied | Not supplied | Independent inspection unavailable |
Copy a prompt for this example
Create a single HTML reading companion for a forty-five-minute lab meeting about an unresolved software comparison. Put the research question and the two illustrative observations at the top: configuration A is 20 ms and B is 28 ms, but their warmup conditions differ. Label the eight-millisecond difference as arithmetic rather than a demonstrated configuration effect.
Provide Question, Evidence, Competing explanations, and Proposed checks buttons. Each opens its own concise reading panel. The Question panel gives the agenda: five minutes of context, ten of evidence, fifteen of explanations, ten of proposed checks, and five of unresolved work.
Evidence must identify missing method versions, warmup counts, replicate information, and raw results. The explanations are E1, configuration contributes, and E2, warmup contributes. Both may contribute and neither is established by the two values. Ask what observation could distinguish or weaken them.
The Proposed checks panel describes P01: compare configurations with the same agreed warmup and inputs, recording versions and protocol. Keep the agreement and later observation explicitly unanswered; P01 is a proposal, not an experiment already performed.
Add one clearly labeled discussion-notes box. Make it clear that this page is a reading and scratch-note companion to the maintained Markdown or DOCX agenda, not a generated DOCX or a research-results repository. Reset returns the Question panel and clears notes.
Use a calm research-notebook layout, readable comparison labels, and stacked sections on phones. Return the complete HTML with its styles and panel controls.
Bring the maintained Markdown or DOCX agenda and raw evidence references to the meeting. A sanitized HTML reading copy can be previewed in NOS alongside the explanation matrix. Close by identifying what the next agreed check must hold constant, and leave owner or date unresolved if nobody has agreed to them.