From 008fc1523a432e34ef10b9166c265ba091b409d3 Mon Sep 17 00:00:00 2001 From: Peter D Bethke Date: Sun, 13 Sep 2026 07:54:36 -0400 Subject: [PATCH] docs: refresh the storage diagram and field note to the live figures The branch has grown since both were written: 104 -> 108 entries, 382 -> 402 KB, 3.6 -> 3.8 KB average, and the 10,000-entry projection from 37 to 38 MB. The diagram's footer now says it was read on 2026-09-13. These numbers are about to be put in front of an audience that will run the query, and a diagram whose count disagrees with what read_json_auto returns is the kind of small gap that costs more credibility than the diagram buys. Co-Authored-By: Claude Opus 5 (1M context) Claude-Session: https://claude.ai/code/session_011NgkB3eLPBhgfqU1XopR1V --- site/public/diagrams/corral-storage.svg | 6 +++--- site/src/content/field-notes/a-ledger-not-a-log.mdx | 4 ++-- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/site/public/diagrams/corral-storage.svg b/site/public/diagrams/corral-storage.svg index 4314d376..cf824c1e 100644 --- a/site/public/diagrams/corral-storage.svg +++ b/site/public/diagrams/corral-storage.svg @@ -117,8 +117,8 @@ Measured, today - 104 entries · 382 KB total · 3.6 KB avg - ≈ 37 MB projected at 10,000 audits + 108 entries · 402 KB total · 3.8 KB avg + ≈ 38 MB projected at 10,000 audits @@ -270,5 +270,5 @@ CORRAL · corralai.dev Cream boxes are model seats · green are code, and ask no model anything · amber-dashed are stored · red outlines are yours, and nothing there happens without you · dotted arrows are read back, never written. - Every number on this diagram was read from the live branch on 2026-09-11: github.com/pdbethke/corralai/tree/corral/ledger/scans · Elastic-2.0 · one Go binary · no hosted tier. + Every number on this diagram was read from the live branch on 2026-09-13: github.com/pdbethke/corralai/tree/corral/ledger/scans · Elastic-2.0 · one Go binary · no hosted tier. \ No newline at end of file diff --git a/site/src/content/field-notes/a-ledger-not-a-log.mdx b/site/src/content/field-notes/a-ledger-not-a-log.mdx index 3503b379..15fff2e9 100644 --- a/site/src/content/field-notes/a-ledger-not-a-log.mdx +++ b/site/src/content/field-notes/a-ledger-not-a-log.mdx @@ -101,8 +101,8 @@ could price an audit from this repository's own history instead of guessing. ## What it costs **It is unindexed, and every query reads every file.** That works because an -entry is one per *audit*, not one per event. Today the whole record is **106 -entries, 386 KB, averaging 3.7 KB each** — about 37 MB projected at ten +entry is one per *audit*, not one per event. Today the whole record is **108 +entries, 402 KB, averaging 3.8 KB each** — about 38 MB projected at ten thousand audits. Benchmarks, by contrast, are one per benchmark per commit per platform, which is the assumption this design rests on, violated. If your facts arrive per-event, this is the wrong shape and a real database is the right one.