Skip to content

Buyer leads: no lead on wording alone when a model is there to judge it - #146

Merged
ralyodio merged 1 commit into
mainfrom
fix/buyer-leads-quality
Oct 7, 2026
Merged

ralyodio merged 1 commit into
mainfrom
fix/buyer-leads-quality

Conversation

@ralyodio

@ralyodio ralyodio commented Oct 7, 2026

Copy link
Copy Markdown
Contributor

First live scan (ThreatCrush, 419 posts) showed three problems:

  • 336 posts past the 60-per-scan model budget kept their wording guess, and
    a game "demo" became a 72-point lead. Unjudged posts now wait at 45 (below
    any floor) and every scan judges the backlog of the last 7 days first.
  • Judging went newest first, so Bluesky's flood took 54 of 60 model calls and
    Reddit got none. Now best wording first within each source, sources in turn.
  • The judge called "learning Wazuh" buyer intent. The prompt now puts
    learning, tutorials, courses and happy users under 50: when unsure, below 50.
    Wording-only scores without a model: only a request for recommendations
    clears 60. Migration 0060 holds the rows 0059 already scored by wording.

🤖 Generated with Claude Code

First live scan (ThreatCrush, 419 posts) showed three problems:
- 336 posts past the 60-per-scan model budget kept their wording guess, and
  a game "demo" became a 72-point lead. Unjudged posts now wait at 45 (below
  any floor) and every scan judges the backlog of the last 7 days first.
- Judging went newest first, so Bluesky's flood took 54 of 60 model calls and
  Reddit got none. Now best wording first within each source, sources in turn.
- The judge called "learning Wazuh" buyer intent. The prompt now puts
  learning, tutorials, courses and happy users under 50: when unsure, below 50.
Wording-only scores without a model: only a request for recommendations
clears 60. Migration 0060 holds the rows 0059 already scored by wording.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@ralyodio
ralyodio merged commit 6d58fde into main Oct 7, 2026
4 checks passed
ralyodio added a commit that referenced this pull request Oct 7, 2026
Asked for a 0-100 number, the model anchored on the floor: after #146 an
EventLog ad scored 80 and a "who wants to be hired" thread and an internship
question scored 60. The judge now answers kind (seeking / problem /
discussion / promo / offtopic) and fit (high / medium / low), and intent is
computed from those (seeking 88, problem 72, discussion 35, promo 5; fit
x1 / x0.85 / x0.5). Promo explicitly covers vendor ads, courses, launches,
job and hiring threads. Migration 0061 sends rows nobody has acted on back to
the backlog, held below the floor, to be judged again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ralyodio added a commit that referenced this pull request Oct 7, 2026
* Buyer leads: the judge classifies, the score is computed

Asked for a 0-100 number, the model anchored on the floor: after #146 an
EventLog ad scored 80 and a "who wants to be hired" thread and an internship
question scored 60. The judge now answers kind (seeking / problem /
discussion / promo / offtopic) and fit (high / medium / low), and intent is
computed from those (seeking 88, problem 72, discussion 35, promo 5; fit
x1 / x0.85 / x0.5). Promo explicitly covers vendor ads, courses, launches,
job and hiring threads. Migration 0061 sends rows nobody has acted on back to
the backlog, held below the floor, to be judged again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: prettier

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
@ralyodio ralyodio mentioned this pull request Oct 7, 2026
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