Evidence in 95 conversations

Source and method

Source: JPC's own count of the transcripts of the 95 items published on /media as of November 1, 2026. Machine-generated transcripts; speaker roles not verified. Method and term list published.

Input files: 193, SHA-256 8da89a5c9574. Generator SHA-256 74a7f1d79ef4. Term list v1.

What a figure is

One published item on /media has one transcript. A figure is the number of those transcripts that contain the word or phrase at least once, and the total number of times it appears.

Text counted

The text after the line TRANSCRIPT in each en.txt file, with any "[mm:ss] Speaker:" label removed. File headers (title, asset id, source file, status, speaker note) are not counted.

Matching

Lower-cased text, curly apostrophes folded to straight ones, case-insensitive regular expression with word boundaries, matched across paragraph breaks. The patterns are listed with each term.

Files read

• scripts/research/corpus-term-frequency.mts (the generator, hashed in generatorSha256)
• lib/media-catalog.ts (records published on /media, sorted by record id)
• lib/media-library.ts and lib/blob/store-config.ts (which records are published, and where each transcript is)
• public/media/accessibility/v1/**/transcript/en.txt
• public/media/accessibility/v1/**/manifest.json (duration_seconds, for total hours only)

Figures

Term list, v1

The list is fixed. The ten terms in the first two groups are listed only if they appear in at least 5 transcripts. The five formal-document terms are listed at any count: they were on the list before that floor removed them, and how rarely they appear is the finding. The acronym SOP is not counted; it does not appear in the transcripts, only the long form does.

• metric, metrics\bmetrics?\b
• checklist, checklists\bchecklists?\b
• spreadsheet, spreadsheets\bspreadsheets?\b
• dashboard, dashboards\bdashboards?\b
• before and after\bbefore[- ]and[- ]after\b|\bbefore/after\b
• living professional record\bliving professional records?\b
• rendering, renderings\brenderings?\b
• operational texture\boperational texture\b
• evidence vault\bevidence vaults?\b
• evidence layer\bevidence layers?\b
• decision record, decision records\bdecision records?\b
• work episode, work episodes\bwork episodes?\b
• process map, process maps\bprocess maps?\b
• standard operating procedure, standard operating procedures\bstandard operating procedures?\b
• meeting notes\bmeeting notes\b

What these counts do not show

• A count shows that a word appears. It does not show that the topic matters.
• The counts cover JPC's own published talk. They are not a sample of any wider group.
• The transcripts are machine-generated, so a word may be mis-transcribed.
• Only the listed words and phrases are counted. A topic can be discussed in other words.
• A mention can be figurative, or part of a rule against doing something. The count does not separate these.
• Several worked examples recur across items, so a total can reflect one repeated example.
• They say nothing about what happens to people who use any method.
• Nothing here describes the Record Builder or anyone's Record.

How the counts were checked

• About 330 matches of the ten terms in the first two groups were read in context. 1 (0.3%) used the word in a different sense, and about 9 (2.7%) used it loosely or figuratively, mostly "dashboard".
• All 8 matches of the formal-document terms were read in context. Each used the term in its ordinary sense.
• The matched lines were checked for the names of private people. None was identified. A check like this can miss lowercase or garbled names.
• A few suspected mis-hearings were found in the text. No one listened to the audio, so a mis-hearing that produces another real word would not be found.

Corpus facts

Published items

95 published items (81 podcast episodes, 11 website videos, 3 website audio)

Transcripts counted

95 of 95

Words in the transcripts

308,866

Total running time

31.6 hours, from the item manifests