Re-render the demonstration with the current test count

The suite is 407 after the award regression tests; the terminal line and the narration follow.

assets/demo/topvenues-demo-v1.14.0.en.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91af927c56003e33319a87a21a43b687d5e20c991aab112527f79fe49abd4cb3
-size 22881100
+oid sha256:59c64def82899e55d86c80f8ffc8e0a0e88ca08b7fb838aab3bd056d71957652
+size 22830955

assets/demo/topvenues-demo-v1.14.0.en.vtt

@@ -36,46 +36,46 @@ It restores the corpus, checks the fingerprint,
 runs the tests, and searches.
 
 8
-00:01:05.875 --> 00:01:13.625
-407 tests, eight environments, every change.
+00:01:05.875 --> 00:01:13.000
+The whole suite, eight environments, every change.
 The same bytes every time.
 
 9
-00:01:14.075 --> 00:01:25.200
+00:01:13.450 --> 00:01:24.575
 Now the questions: 15,286 papers,
 20 declared venues, 2019 to 2026.
 
 10
-00:01:25.650 --> 00:01:35.350
+00:01:25.025 --> 00:01:34.725
 Search happens inside the file: an index finds,
 a relevance formula ranks. No internet, no model.
 
 11
-00:01:35.800 --> 00:01:42.975
+00:01:35.175 --> 00:01:42.350
 The scope is stated above the results:
 what was searched, not just what was found.
 
 12
-00:01:43.425 --> 00:01:51.275
+00:01:42.800 --> 00:01:50.650
 The same query on the command line,
 with BibTeX you can paste into the paper.
 
 13
-00:01:51.725 --> 00:02:01.650
+00:01:51.100 --> 00:02:01.025
 It also says where it is silent: 892 records
 carry no abstract, declared source by source.
 
 14
-00:02:02.100 --> 00:02:12.600
+00:02:01.475 --> 00:02:11.975
 With the list fixed, things become measurable:
 3 in 10 top-4 papers were already on arXiv.
 
 15
-00:02:13.050 --> 00:02:24.200
+00:02:12.425 --> 00:02:23.575
 One question separates preprints by 16.5×,
 replicated on a second cohort.
 
 16
-00:02:24.650 --> 00:02:31.300
+00:02:24.025 --> 00:02:30.675
 Declare the list. Freeze it. Cite it.
 Code, corpus and papers are open.

assets/demo/topvenues-demo-v1.14.0.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b4c8997b146dfe41c21be0eba74f509fcd696bb2367e00c4d3fa74d8dc016137
-size 22225315
+oid sha256:b7787ba9d70b884fac8464da3333b9599aa79f8fb424b787ba8ed779018d9701
+size 22210478

assets/demo/topvenues-demo-v1.14.0.pt-BR.mp4

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b7a7aa961d15ec81d249e72837176a126f1fb20905433390cd40dd849064899
-size 22905622
+oid sha256:e11618b5535970ca98fa57529a13cd3416f86c228373249725097b2984b898e9
+size 22860988

assets/demo/topvenues-demo-v1.14.0.pt-BR.vtt

@@ -36,46 +36,46 @@ Ele restaura o corpus, confere a impressão digital,
 roda os testes e busca.
 
 8
-00:01:05.875 --> 00:01:13.625
-407 testes, oito ambientes, a cada mudança.
+00:01:05.875 --> 00:01:13.000
+A suíte inteira, oito ambientes, a cada mudança.
 Sempre os mesmos bytes.
 
 9
-00:01:14.075 --> 00:01:25.200
+00:01:13.450 --> 00:01:24.575
 Agora as perguntas: 15.286 artigos,
 20 veículos declarados, de 2019 a 2026.
 
 10
-00:01:25.650 --> 00:01:35.350
+00:01:25.025 --> 00:01:34.725
 A busca acontece dentro do arquivo: um índice acha,
 uma fórmula ordena. Sem internet e sem modelo.
 
 11
-00:01:35.800 --> 00:01:42.975
+00:01:35.175 --> 00:01:42.350
 O escopo aparece acima dos resultados:
 o que foi buscado, não só o que foi achado.
 
 12
-00:01:43.425 --> 00:01:51.275
+00:01:42.800 --> 00:01:50.650
 A mesma consulta na linha de comando,
 com BibTeX para colar direto no artigo.
 
 13
-00:01:51.725 --> 00:02:01.650
+00:01:51.100 --> 00:02:01.025
 Ele também diz onde se cala: 892 registros
 sem resumo, declarados fonte por fonte.
 
 14
-00:02:02.100 --> 00:02:12.600
+00:02:01.475 --> 00:02:11.975
 Com a lista fixa, dá para medir:
 3 em cada 10 artigos do top-4 já estavam no arXiv.
 
 15
-00:02:13.050 --> 00:02:24.200
+00:02:12.425 --> 00:02:23.575
 Uma pergunta separa preprints em 16,5×,
 replicado numa segunda coorte.
 
 16
-00:02:24.650 --> 00:02:31.300
+00:02:24.025 --> 00:02:30.675
 Declare a lista. Congele. Cite.
 Código, corpus e artigos estão abertos.