Aggregate · proof, quantified
Stats
Aggregate stats across every published reproduction. Spend figures cover the 247 reproductions with measured cost; counts cover all 276.
276
Published
247 /276
With measured cost
$4.30
Median spend
$2673.68
Total spent
Min
$0.04
P90
$23.80
Max
$369.27
Cost distribution
247 reproductions with measured cost, bucketed by total spend.
<$0.10
1
$0.10–$0.50
22
$0.50–$1.00
11
$1.00–$2.00
32
$2.00–$5.00
67
≥$5.00
114
Spend by agent and model
Total dollars charged to each agent, split by LLM. The sum may exceed the measured-cost total because each reproduction's spend is fully attributed.
repro
$1935.21
vuln_variant
$588.99
judge
$84.38
support
$34.47
coding
$23.93
claim_matcher
$2.33
hypothesis_generator
$1.66
discovery
$1.51
learning_policy
$1.19
accounts/fireworks/models/glm-5p2
accounts/fireworks/models/glm-5p3
accounts/fireworks/models/kimi-k2p5
accounts/fireworks/models/kimi-k2p6
accounts/fireworks/models/kimi-k2p7-code
accounts/fireworks/models/kimi-k3
accounts/fireworks/routers/glm-5p2-fast
claude-opus-4-7
gpt-5.1-codex
gpt-5.2-codex
gpt-5.4-mini
gpt-5.4-mini-2026-03-17
gpt-5.5
gpt-5.5-2026-04-23
gpt-5.6-sol
By ecosystem
Count, median cost, and median duration per package ecosystem.
| Ecosystem | Count | Median cost | Median duration | Share |
|---|---|---|---|---|
| github | 80 | $7.01 | 45m 2s | |
| npm | 33 | $1.05 | 16m 25s | |
| pip | 28 | $2.45 | 17m 47s | |
| composer | 11 | $0.55 | 13m 7s | |
| go | 11 | $1.92 | 27m 11s | |
| c | 7 | $1.33 | 29m 29s | |
| maven | 6 | $7.61 | 43m 30s | |
| other | 5 | $5.41 | 46m 36s | |
| Maven | 3 | $10.78 | 48m 55s | |
| PyPI | 3 | $6.76 | 41m 3s | |
| generic | 3 | $1.64 | 21m 31s | |
| source | 3 | $1.68 | 16m 28s | |
| Composer | 2 | $12.42 | 61m 3s | |
| Go | 2 | $14.78 | 26m 51s | |
| firmware | 2 | $18.88 | 61m 51s | |
| linux | 2 | $6.90 | 67m 57s | |
| unknown | 2 | $7.48 | 79m 4s | |
| vendor | 2 | $15.30 | 104m 40s | |
| CPAN | 1 | $4.14 | 17m 47s | |
| GitHub / Node.js web application | 1 | $1.48 | 14m 14s | |
| Go module | 1 | — | 38m 55s | |
| Joomla extension | 1 | $10.73 | 77m 51s | |
| Ruby | 1 | $1.67 | 18m 33s | |
| Ruby on Rails self-hosted application, also shipped as Docker image pglombardo/pwpush | 1 | $2.32 | 21m 33s | |
| WordPress plugin (hosted on WordPress.org SVN, not GitHub) | 1 | $7.05 | 22m 56s | |
| cargo | 1 | $4.18 | 33m 23s | |
| cpp | 1 | — | 47m 53s | |
| docker | 1 | $3.15 | 44m 34s | |
| gnu | 1 | — | 35m 50s | |
| joomla | 1 | $18.76 | 107m 17s | |
| nodejs | 1 | $3.26 | 47m 12s | |
| npm / JavaScript (Node.js, Express) | 1 | $2.01 | 12m 11s | |
| other (commercial, Java-based server application) | 1 | $37.97 | 94m 25s | |
| pip (per GitHub advisory) | 1 | — | 8m 19s | |
| product | 1 | $6.41 | 122m 25s | |
| pypi | 1 | — | 4m 50s | |
| rubygems | 1 | $0.67 | 10m 15s | |
| rust | 1 | $0.27 | 10m 6s | |
| standalone application (Python/FastAPI) | 1 | $2.02 | 14m 56s | |
| wordpress | 1 | $7.43 | 20m 35s | |
By severity
Count and median cost per CVSS severity bucket.
critical 112 · median $5.39
high 124 · median $3.63
medium 30 · median $2.69
low 8 · median $6.01