Sibyl Memory — independent beta suites: scale, retention, answering (June 2026)

Independent closed-beta tester runs that scaled the corpora 10x and measured
whether retrieval and answering hold as the dataset grows. Synthetic data; the
tester's handle and OS paths were redacted to "tester". The runners read API
keys from the environment; none are included.

Contents:
  scale10x-retention-report.md, dense10x-retention-report.md
      10x and dense-10x write/retention checkpoints (100% retained, DB stays small)
  run294-chronology-retention.(md, raw.json)  chronology volume retention source
  run295-chronology-answering.(md, raw.json)  12/12 retrieval contained the answer
  run296-strict-facts-answering.(md, raw.json)    12/12 retrieval contained the answer
  run297-scale10x-answering.(md, raw.json)    37/38 retrieval at 11,520 writes
  noncanonical-business-memory-report.md + .raw.json + -suite.py
  segment-evidence-tracks-report.(md, raw.json)
      segment evidence paths: top-8 keyword 233/240, full company context 240/240

Write-up: https://blog.sibylcap.com/beta-analysis
Sibyl Labs, LLC
