docs: enact RFC 0012 — §2 pillar-#2 is a logical 50–200×, not byte-level

CLAUDE.md

@@ -51,9 +51,16 @@ decision, not a PR-level one.
    and page indexes. Query performance comes from skipping row groups via
    footer reads, not from scanning.
 2. **Drain-derived online template mining.** Log lines collapse to
-   `(template_id, params)` at ingest time. This is where the 50–200×
-   compression comes from — before any byte-level codec runs. Correctness
-   of this layer is the single biggest engineering risk in the project.
+   `(template_id, params)` at ingest time — a **logical** 50–200×
+   reduction (many near-identical lines become rows keyed by one small,
+   stable `template_id`). That reduction is what lets a selective query
+   read a handful of row groups instead of scanning the corpus, so the
+   payoff is **query pruning** (pillar #1's footer-skip; benchmark gates
+   B1/B2), **not** fewer on-disk bytes than a byte codec — RFC 0011
+   showed a whole-stream codec captures the same redundancy, so the
+   on-disk-compression-vs-zstd ratio (A1) is a recorded diagnostic, not a
+   gate. Correctness of this layer is the single biggest engineering risk
+   in the project.
 3. **DataFusion as the query engine.** We do not write a vectorised
    execution engine. We hand DataFusion logical plans and let it work.
 
@@ -394,11 +401,15 @@ the wrong thing.
 
 ---
 
-*Last updated: 2026-05-13. Original draft 2026-04-23. 2026-04-26 revision
+*Last updated: 2026-06-14. Original draft 2026-04-23. 2026-04-26 revision
 aligned §7 with the present-day repo and added §6.7 + an mdBook entry to
 §6.6 (commits 1c806f5..84eed86). 2026-05-13 revision adds the "Tests are
 specifications, not friction" bullet to §6.2, codifying a discipline
 called out by Dave Farley as a recurring AI-assisted-development failure
 mode (informal `meta:` RFC waiver per maintainer discussion of the same
-date; precedent: b50067d). This document is load-bearing; further changes
-require a `meta:` RFC and majority maintainer approval.*
+date; precedent: b50067d). 2026-06-14 revision rewords §2 pillar #2 so the
+50–200× reads as a **logical** reduction (query pruning, benchmark gates
+B1/B2) rather than an on-disk-bytes-beat-the-codec claim, per **RFC 0012**
+(maintainer-approved `meta:` RFC) propagating RFC 0011's A1 demotion.
+This document is load-bearing; further changes require a `meta:` RFC and
+majority maintainer approval.*

README.md

@@ -21,8 +21,9 @@ parts:
 - **Apache Parquet** — columnar storage, with predicate pushdown via
   row-group statistics, bloom filters, and page indexes.
 - **A Drain-derived online template miner** — collapses log lines to
-  `(template_id, params)` at ingest time, before any byte-level codec
-  runs.
+  `(template_id, params)` at ingest time: a *logical* 50–200× reduction
+  that drives query pruning (not a bytes-vs-codec claim; see
+  [RFC 0011](docs/rfcs/0011-a1-rescope.md)).
 - **Apache DataFusion** — a production-ready vectorised SQL execution
   engine. We hand it logical plans; it does the work.
 

docs/benchmarks.md

@@ -121,9 +121,12 @@ codecs run.
 - **Falsifier**: if any representative corpus yields ≤ 2× improvement
   over zstd-alone, the template-mining pillar is not pulling its
   weight on that class of logs. Open an RFC.
-- **Why this bar**: the project's headline claim (§2, CLAUDE.md) is
-  "50–200× over raw, ≥ 5× over a competent byte codec." Below 3× the
-  headline is dishonest.
+- **Why recorded** (diagnostic, not a bar): `CLAUDE.md` §2 pillar #2
+  describes a **logical** 50–200× reduction (lines → `(template_id,
+  params)`) whose payoff is query pruning (B1/B2), *not* on-disk bytes vs
+  a byte codec. A1 tracks the on-disk ratio as the columnar queryability
+  premium + a codec-regression guard; RFC 0011 (`accepted`) demoted it
+  from a gate to this diagnostic.
 
 ### A2 — Bytes per line, amortised
 

docs/rfcs/0001-template-miner.md

@@ -80,8 +80,10 @@ rather than a property-test exception `[§3.3]`, bounds parameter
 values at 256 B with overflow to a side `body` column `[§3.2]`, and
 tracks template structural changes via a monotonic `template_version`
 so that schema drift across deploys is a first-class query rather than
-a silent count drop `[§3.5]`. The compression target is 50–200× over
-raw bytes before any byte-level codec runs.
+a silent count drop `[§3.5]`. The 50–200× figure is a **logical**
+reduction (lines → `(template_id, params)`), realised as query pruning
+(gates B1/B2), not on-disk bytes versus a byte codec — see RFC 0011, which
+demoted the compression-vs-zstd ratio (A1) to a diagnostic.
 
 ## 2. Motivation
 

docs/rfcs/0012-claude-md-pillar-2-wording.md

@@ -1,7 +1,7 @@
 ---
 rfc: 0012
 title: "meta: CLAUDE.md §2 pillar-#2 wording — template mining's 50–200× is a logical reduction, not on-disk bytes"
-status: drafted
+status: accepted
 author: Jens Holdgaard Pedersen <jens@holdgaard.org>
 drafting-assistance: Claude
 created: 2026-06-14
@@ -11,6 +11,15 @@ superseded-by: —
 
 # RFC 0012 — meta: CLAUDE.md §2 pillar-#2 wording
 
+> **Status note. `accepted` (2026-06-14, maintainer-approved + enacted.)**
+> The `CLAUDE.md` §2 pillar-#2 reword (§3.1) and all three coupled
+> reconciliations (§3.2 — `benchmarks.md` §2, `README.md`, RFC 0001 §1)
+> were applied in the enacting PR; RFC 0001's accepted prose took the
+> recommended factual reword (with the RFC 0011 pointer). The §7 footer
+> changelog line + `Last updated` bump landed in the same diff. This
+> meta-RFC required majority maintainer approval per `CLAUDE.md`'s footer;
+> that gate is satisfied.
+
 > **This is a `meta:` RFC.** It proposes a change to `CLAUDE.md`, which
 > its own footer declares load-bearing: *"This document is load-bearing;
 > further changes require a `meta:` RFC and majority maintainer