memory: snapshot — greybox build complete, playtest pending

First /snapshot: seeds persistent-memory.md (+ one detail file for the freeze-meter variant decision), amends CLAUDE.md with the persistent- memory pointer section.
2026-08-07 12:02:51 -07:00
parent 1976149ce5
commit 45c0c78481
3 changed files with 151 additions and 0 deletions
@@ -0,0 +1,58 @@
+# 2026-08-07 — Freeze charge meter adopted as the card-time playtest variant
+
+**Status: playtest variant, NOT canon.** Ratification into the advisor
+repo's CONCEPT.md is deferred until the operator's falsification-playtest
+verdict — tracked at commit `73f84e7` (whose message records the pending
+state) and the advisor repo's CONCEPT.md § Open questions #0 (the greybox
+work tracker).
+
+## The mechanic
+
+A "Freeze" charge (meter) grows while the hand is closed, drains while
+open. Its level maps to bands, best to worst:
+
+- **FROZEN** (level ≥ 45%): true pause (`SceneTree.paused`).
+- **THAW** (≥ 18%): slow-mo ramping back to full speed as it drains, on
+  an acceleration curve (`thaw_curve` 2.2 — hangs slow, then rushes).
+- **REALTIME** (≥ 10%): hand open, world at full speed.
+- **UNAVAILABLE** (below): the hand will not open at all; a drained-open
+  hand force-closes.
+
+Opening with a full meter buys the whole ladder (~1.65s freeze → ~0.8s
+thaw → ~0.5s realtime at defaults: cap 3.0s, drain 1.0/s, regen 0.4/s).
+Closing the hand spools time back up over `release_thaw_s` (0.5s) instead
+of snapping. All rates/bands are `@export` knobs; M flips live to the
+original unlimited-toggle variant for A/B.
+
+## Why (KB-grounded pressure test, 2026-08-07)
+
+- **Kills a latent dominant strategy**: unlimited freeze made "open the
+  hand as a free tactical pause" always-optimal (Schell: rework rules so
+  a dominant strategy stops being dominant; aligns with CONCEPT.md's
+  anti-dominant-strategy posture). The empty-meter lockout also closes
+  the Space-spam refreeze exploit found during the pressure test.
+- **Tension cycle**: freeze→thaw→realtime is Schell's
+  rising-tension-then-release loop in miniature, inside every hand-open.
+- **Temporal triangularity**: play one card early (safe) vs chain deeper
+  into the thaw (greedy) — stacks with belt-depth pricing post-greybox.
+- **Difficulty/accessibility axis**: meter size / regen / thaw length are
+  one-scalar difficulty knobs that never touch enemy content — can't
+  collide with the invincibility laws. Generous-to-infinite freeze is the
+  accessibility floor.
+- **Honesty note**: the KB surfaced nothing directly on decision timers /
+  time pressure (retrieval miss) — grounding is flow/tension/triangularity/
+  dominant-strategy material, adjacent but not a direct verdict.
+
+## Canon tension to watch in playtest
+
+"Attention is spent ON cards, not during them" (pillar 1, gentlest action
+skill): a tight freeze window would turn the hand into a reflex test. The
+generous default + continuous thaw (never a cliff) is the mitigation; the
+ramp prices dithering and deep chains, not basic use.
+
+## Evidence chain
+
+Pressure test + operator "A for now" directive: session 2026-08-07.
+Implementation: `73f84e7` (meter + bands), `3d2410c` (release ramp +
+thaw curve). Operator early verdict: "that DOES feel very good", "ramp
+feels good too… starting to feel game-like".