Query proof

Chart a community's drift along a meaning axis with SQL

A Scry query pattern that mints two poles as vectors, takes the balanced axis between them, and averages every LessWrong post's projection quarter by quarter.

Answer shape

Two embed calls mint the poles; one SQL query joins chunk embeddings to real post timestamps and returns the community's lean along the axis, quarter by quarter.

Sources

  • LessWrong

Methods

  • compositional embeddings
  • contrast axis
  • source-native JOIN
  • time series

SQL

Run this through POST /v1/scry/query after inspecting GET /v1/scry/schema for current live fields.

-- Mint the poles once, then query:
-- POST /v1/scry/embed {"text": "we will solve alignment", "name": "optimism"}
-- POST /v1/scry/embed {"text": "we are not going to make it", "name": "doom"}
SELECT toStartOfQuarter(p.original_timestamp) AS quarter,
       avg(scry_cosine_similarity(e.embedding_voyage4,
           scry_contrast_axis_balanced(@optimism, @doom))) AS lean,
       uniq(e.post_key) AS posts
FROM embeddings.forum_posts AS e
JOIN forums.posts AS p ON p.post_key = e.post_key
WHERE e.model_name = 'voyage-4-lite' AND p.source = 'lesswrong'
GROUP BY quarter
ORDER BY quarter
LIMIT 100;

Verification notes

  • NoteVectors are named SQL values, minted from any text via POST /v1/scry/embed.
  • Notescry_contrast_axis_balanced(@optimism, @doom) is the axis between the two poles.
  • NoteEach quarter keeps uniq(post_key), so the sample size behind every point stays visible.