Files
Jyotisha/docs/research
Jesse_ChenandClaude Opus 5.5 bcf0c86feb research(rectification): typed events scored as answered probes — no benefit (BUG-1089)
v4 open holdout, ±10/±30/±60, raw and percent priors: only 4/9/14 of 57
day-precision training events split candidates; widths unchanged, top-1
drops. Narrowed year blocking (R3) mixed. All arms no_benefit; no
implementation brief recommended. Two runs byte-identical.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
2026-09-29 10:06:13 +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  • it is superseded by a newer audit, dashboard, or implementation note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