Commit c7c4ccb
committed
docs: the whole-history sweep this corpus has not done, and what it costs
Every one of the 2,225 rows reached the classifier because a keyword or an
advisory found it first. There are no rows drawn from whole commit histories,
the curated set covers 0.291% of the 595,966-commit history, and the per-client
coverage spread is 7.3x — which is the same confound the sampling work already
ran into. The sister wallet corpus moved from this exact position to reading
histories end to end and measured the cost of the keyword route at 81% of fixes
missed, so the gap here is probably comparable.
docs/TODO.md sequences the work and puts numbers on it. The screen has to be
scored against the 300-commit external validation set before anything is swept:
at recall 32.7% and an 18.7% positive rate on random commits, sweeping now would
waste most of what it costs. That check is 0.4-0.7M tokens; the first client
(grandine) is 5M; the four smallest are 96M batched, 171M as one call per commit.
Measured from the wallet caches: 2,150 input and 77 output tokens per judgement,
57% of the input being a static prompt that batching would amortise.
It also records the traps the wallet sweep hit after publishing, so they get
built in rather than discovered: reachability and squash-duplicate flags (892 of
5,457 rows never shipped, 475 were the same fix twice), release-tag verification
for any row shown as an example (3 of the first 5 candidates failed it), and no
patch-id backport detection, which did not survive testing as a signal.1 parent 5249b6e commit c7c4ccb
2 files changed
Lines changed: 142 additions & 0 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
117 | 117 | | |
118 | 118 | | |
119 | 119 | | |
| 120 | + | |
| 121 | + | |
| 122 | + | |
| 123 | + | |
| 124 | + | |
| 125 | + | |
| 126 | + | |
| 127 | + | |
| 128 | + | |
| 129 | + | |
| 130 | + | |
| 131 | + | |
120 | 132 | | |
121 | 133 | | |
122 | 134 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
| 13 | + | |
| 14 | + | |
| 15 | + | |
| 16 | + | |
| 17 | + | |
| 18 | + | |
| 19 | + | |
| 20 | + | |
| 21 | + | |
| 22 | + | |
| 23 | + | |
| 24 | + | |
| 25 | + | |
| 26 | + | |
| 27 | + | |
| 28 | + | |
| 29 | + | |
| 30 | + | |
| 31 | + | |
| 32 | + | |
| 33 | + | |
| 34 | + | |
| 35 | + | |
| 36 | + | |
| 37 | + | |
| 38 | + | |
| 39 | + | |
| 40 | + | |
| 41 | + | |
| 42 | + | |
| 43 | + | |
| 44 | + | |
| 45 | + | |
| 46 | + | |
| 47 | + | |
| 48 | + | |
| 49 | + | |
| 50 | + | |
| 51 | + | |
| 52 | + | |
| 53 | + | |
| 54 | + | |
| 55 | + | |
| 56 | + | |
| 57 | + | |
| 58 | + | |
| 59 | + | |
| 60 | + | |
| 61 | + | |
| 62 | + | |
| 63 | + | |
| 64 | + | |
| 65 | + | |
| 66 | + | |
| 67 | + | |
| 68 | + | |
| 69 | + | |
| 70 | + | |
| 71 | + | |
| 72 | + | |
| 73 | + | |
| 74 | + | |
| 75 | + | |
| 76 | + | |
| 77 | + | |
| 78 | + | |
| 79 | + | |
| 80 | + | |
| 81 | + | |
| 82 | + | |
| 83 | + | |
| 84 | + | |
| 85 | + | |
| 86 | + | |
| 87 | + | |
| 88 | + | |
| 89 | + | |
| 90 | + | |
| 91 | + | |
| 92 | + | |
| 93 | + | |
| 94 | + | |
| 95 | + | |
| 96 | + | |
| 97 | + | |
| 98 | + | |
| 99 | + | |
| 100 | + | |
| 101 | + | |
| 102 | + | |
| 103 | + | |
| 104 | + | |
| 105 | + | |
| 106 | + | |
| 107 | + | |
| 108 | + | |
| 109 | + | |
| 110 | + | |
| 111 | + | |
| 112 | + | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
| 119 | + | |
| 120 | + | |
| 121 | + | |
| 122 | + | |
| 123 | + | |
| 124 | + | |
| 125 | + | |
| 126 | + | |
| 127 | + | |
| 128 | + | |
| 129 | + | |
| 130 | + | |
0 commit comments