Skip to content

LODESTONE

A faint compass for a from-scratch search: nudge it to commit the rare pieces early, where they're needed. It doesn't raise the ceiling; it makes the search reliably reach the top of its own range.

by Raphaël Anjou

experimentsolvermeasuredplateaued451/480matched edgesBuild boards upReduce the searchLearned guidanceUpdated 2026-07-21
Reproducestochastic — won't reproduce exactly; the board is verifiablere-verifies a stored boardBudget: not logged (exploratory run, not the standardized single-core bench)
Pipeline
  1. 1
    beam producernot yet written up

    Build from empty; a scarce-demand weight breaks ties as a pure tiebreak, never the objective

    carries: Rare-colour prior: a per-piece weight for how often a piece serves a scarce north-west demand

Complexity
Time
prior build O(B) over the corpus; beam construction as PRIOR/KEYRING, plus a tiebreak lookup
Space
one scarce-demand weight per piece (a 256-vector)

The prior is a per-piece scalar; it costs nothing at search time beyond a tiebreak, which is exactly why over-weighting it (making it part of the objective) is what breaks it.

Hardware & run
Native runCPU only
Cores
8
RAM
16 GiB
GPUs
0
CPU
Apple M1
Machine
MacBook (Apple M1, 8 cores)
Budget
not logged (exploratory run, not the standardized single-core bench)
Start
stochastic; see the run's repro command
Verify the stored board

stochastic — won't reproduce exactly; the board is verifiable

This verifies the board, not the search that produced it.

just research-record-boards
Code & data on GitHub

Keep exploring

Referenced by

Page sourceView as Markdown