Skip to content

feat(marketing): an unnamed speaker is written in the third person - #77

Merged
jgleal merged 2 commits into
mainfrom
feat/unnamed-speaker-third-person
Oct 9, 2026
Merged

jgleal merged 2 commits into
mainfrom
feat/unnamed-speaker-third-person

Conversation

@jgleal

@jgleal jgleal commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

What happened

In orbitant-employents, Mercuria holds the draft until somebody names whoever speaks in an interview or talk (ADR-026). The owner ruled (2026-09-26, reconfirmed 2026-10-08) that the ask must not block: write first, ask once at the end, then place the name with exact substitutions and never rewrite the piece.

That work was parked (weorbitant/orbitant-employents#137) for one measured reason: with nobody named, Sonnet writes the speaker's story in the first person ("mi equipo", "nuestro almacén", ADR-026, 2026-09-15), and then a name has nowhere to go. On a third-person draft ("la persona entrevistada…") placing the name worked. How to write about somebody not named yet is a writing rule, so under ADR-020 it belongs here, not in the harness.

What changes

  • blog-post-create: new subsection under Input, "Spoken material whose speaker is not named yet". Third person for everything the speaker lived; one neutral phrase used the same way throughout ("la persona entrevistada", "la persona que da la charla"); no invented name, role, gender or company; neither the signer nor the institutional "nosotros" takes their story.
  • x-thread, x-post: the same rule beside "Somebody else's words" ("the person interviewed", "the speaker").
  • linkedin-post: the same rule in Input, for a source resting on an unnamed speaker.
  • yt-description: the speaker line uses the neutral phrase instead of a name nobody gave.
  • Version: marketing 1.9.0 -> 1.10.0 (MINOR, behaviour change), meta 1.4.0 -> 1.5.0, each changed skill's frontmatter bumped.

Talks are covered as well as interviews: a recorded Knowledge Sharing reaches the same path.

Measured

Measured from orbitant-employents (weorbitant/orbitant-employents#269) with this branch's built package swapped into node_modules locally (nothing published, no pin changed), Sonnet as the writer, five blog-post-create drafts from an unnamed interview (x3) and talk (x2):

  • neutral phrase used 5/5, speaker's story in the first person in the body 0/5 (one slip in a slug: "asi-bajamos-de-40-a-7-minutos");
  • a later name placed by grounded substitution 5/5, every body reference;
  • baseline on 1.9.0 with the same harness fact line, 1 rep: no first-person claim either, but "ella" and varied phrasing, so the name found 2 places and every "ella" stayed. The rule is what gives a name somewhere to go.

Not measured: linkedin-post, x-thread, x-post and yt-description outputs.

Checks

npm run check green (lint, validate, tests, package build: @orbitant/brain-marketing@1.10.0, @orbitant/brain@1.5.0).

Not merged and not published: a release tag is needed before the employents PR can pin 1.10.0.

jgleal added 2 commits October 8, 2026 09:27
When the material is somebody speaking (an interview, a recorded talk, a KS
session) and nothing says who they are, the writing skills had no rule for it.
Measured in orbitant-employents on 2026-09-15 (ADR-026): the same interview
transcript with a line naming the interviewee came out as a third-party case
study; without it, it came out in the first person ("mi equipo", "nuestro
almacén"), telling another company's story as Orbitant's.

Mercuria is about to stop holding the draft until somebody names the speaker:
it will write first and ask at the end, then place the name with exact
substitutions and never rewrite the piece. That only works if the piece has a
place for the name. On a third-person draft ("la persona entrevistada") it
worked on the proxies measured; on a first-person one there was nowhere to put
it.

So the rule goes where ADR-020 says writing rules live, in the brain:

- blog-post-create: a subsection under Input. Third person for everything the
  speaker lived, one neutral phrase used the same way throughout ("la persona
  entrevistada", "la persona que da la charla"), no invented name, role,
  gender or company, and neither the signer nor the institutional "nosotros"
  takes their story.
- x-thread, x-post: the same rule beside "Somebody else's words", in English
  by default.
- linkedin-post: the same rule in Input, for a source that rests on an
  unnamed speaker.
- yt-description: the speaker line uses the neutral phrase instead of a name
  nobody gave.

Talks are covered as well as interviews, since a recorded session reaches the
same path.
MINOR, because five skills behave differently for an unnamed speaker.
Everything moves together, as validate-versions.cjs and this repo's precedent
require: plugin.json, marketplace.json, README's table and install line,
.claude/CLAUDE.md's tree and Current State table, the meta package (1.4.0 ->
1.5.0), and each changed skill's own frontmatter.
@jgleal
jgleal requested a review from a team as a code owner October 8, 2026 07:27
@jgleal
jgleal merged commit 021f26d into main Oct 9, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant