Files
Jyotisha/docs/research
jesse-uxandGrok (agent loop) 8e0503f9c0 Speed up rectification scoring without changing results.
Cache work that depends only on the candidate minute, drop unread response fields, and reject windows above 241 candidates before any chart is built. Eleven canonical scenario hashes match 77057a819. Sealed and reported-offset replays match the 2026-10-08 records (changed_trial_count 8 and 581) and are stored as scoring_perf_2026_10_13.

Co-Authored-By: Grok (agent loop)
2026-10-14 02:06:14 +08:00
..

Research Workspace

This directory stores research notes, audits, benchmark packets, and active adjudicator iteration records for the Jyotish project.

Layout

  • active/
    • reserved for current working research fronts that directly support the next implementation cycle
  • archive/
    • historical round notes, sidecar work orders, and local draft research that should not clutter normal implementation work
  • top-level docs/research/*.md
    • current canonical research notes that are still referenced by active engineering work

What Stays Top-Level

Keep a research file at the top level only if at least one of the following is true:

  • it is referenced by an active implementation spec
  • it is referenced by tests, workflow docs, or benchmark dashboards
  • it captures a currently active adjudicator audit or closure board

What Moves To Archive

Move a research file into archive/ when:

  • it is a completed round log or sidecar work order
  • it is a local exploration draft with no active code dependency
  • it is superseded by a newer audit, dashboard, or implementation note

Current Working Rule

When adding new research notes:

  1. put active, decision-driving notes at the top level
  2. move noisy round-by-round drafts into archive/
  3. avoid storing transient terminal captures or scratch outputs here