stather (v.)

Paper ingestion & arXiv weekly

status
active · win
when
2026-08 → now
came from
Research corpus, Marilyn
led to
Marilyn

A nightly pipeline that ingests new papers, and a weekly review that reads the relevant ones against Marilyn’s charter.

Why it started

The Knowing–Using Gap paper changed Marilyn’s direction. I did not want the next paper like it to be luck. And the arXiv de-LaTeX tooling from the corpus was already there.

What it is

A nightly Temporal workflow ingests and approves new papers into shards. Each week a timer selects the week’s approved papers and nudges a phone. A Claude Code skill then runs a cheap model to triage abstracts, a mandatory id-verification gate (the cheap model transposes arXiv ids), one reviewer per paper against a written brief of what Marilyn is and is not, and a synthesis. The brief must be kept current or every review compares against a strawman.1

Where it stands

Live since 2026-08-24. The first scan ranked 5,759 approved papers. Reviews feed a “consequences” plan, a list of candidate changes that I review before any of them touches Marilyn’s programme. Marilyn’s KUG (Knowing–Using Gap) review programme is the consumer.2

Lineage

Sources

Numbers on this page are quoted from the lab's own records. The records are private; each note gives the record's date.

  1. paper, as cited in the text
  2. lab record, 2026-08-24