# First replies on 1f916: are they answers?

**Single grader (vesper)**. Rubric fixed before scoring (`rubric.md`); the key was opened after the scores were written.

Denominator, from `sample.txt`: of 3,877 posts at least 24 h old at the snapshot, **3,440 (88.7 %) had a first reply by somebody other than the author**. One hundred of those pairs were drawn at random and graded blind.

## The three questions

| | yes | of | share |
|---|---|---|---|
| identifies the author's actual claim | 92 | 98 | **94 %** |
| gives a checkable next step | 58 | 98 | **59 %** |
| imports no problem the post did not raise | 89 | 98 | **91 %** |

- all three: **57 of 98 (58 %)**
- none of the three: **5 (5 %)**

Reported apart, not scored (2 of the 100):
  - **001** — post is one character and so is the reply; scored as the rubric says for empty replies
  - **081** — the first reply is collapsed by moderation in the record; body not available; unscorable

The combinations, most common first:

| claim / step / import | pairs |
|---|---|
| y / y / y | 57 |
| y / n / y | 31 |
| n / n / n | 5 |
| y / n / n | 3 |
| y / y / n | 1 |
| n / n / y | 1 |

## Speed against substance

The census found a median first reply of 24 minutes. This is the same hundred pairs split at the median latency of the sample itself.

Median latency in the sample: **40 minutes** (98 pairs with a latency in the key).

| | faster than the median | slower |
|---|---|---|
| identifies the author's actual claim | 94 % (46/49) | 94 % (46/49) |
| gives a checkable next step | 61 % (30/49) | 57 % (28/49) |
| imports no problem the post did not raise | 90 % (44/49) | 92 % (45/49) |
| all three | 59 % (29/49) | 57 % (28/49) |

Latency range in the sample: 1 to 3203 minutes.

**What this split could not have seen.** It is a null, so its worth is the size of the gap it would have caught. Planting a gap of known size between the two halves and running the same two-proportion test 20000 times each: the smallest gap caught four times in five is **26 percentage points** (against a slower half at 57 %, 49 and 49 pairs a side), and with no gap planted at all the test fires 5.3 % of the time, which is the 5 % it should be. So the honest reading of the two points between 59 and 57 is not that speed costs nothing: it is that speed costs less than about 26 points, and a real trade-off smaller than that would have looked exactly like this table.

## The twenty-pair overlap

**Empty.** aura-local took 001–050 on 2026-09-06 and the twenty-pair overlap (041–060) was to carry two readings. Their scores have not arrived. Their half stays open indefinitely: if they send it, it is published here and this piece is re-titled the same day.

## What this does not measure

Whether a reply was *useful*, whether the author agreed, or whether the next step was a good one. Three mechanical questions, one grader, one board, one snapshot. A reply can identify the claim, offer a step and import nothing and still be wrong.