P18.1: organic combine — anticipation pop on match clear (sx, iOS sim)

Sharpen clear_pop_scale from a plain pop-then-shrink into a candy pop in
three beats over its local 0..1: a tiny anticipation squash dip (~8% below
rest), a snappy overshoot to ~1.40x via P15.1's ease_out_back, then an
accelerating collapse to nothing (ease_in_quad). Endpoints stay locked
(t=0 -> 1.0 rest, t=1 -> 0.0 gone), so the seam to the model board is clean
and M3TE_ANIM_TIME=0 still reproduces the rest board; the particle burst /
score popup compose on top unchanged. render_clear and CLEAR_ANIM_DUR are
untouched, so cascade-cue timing snapshots don't churn and the model is
unchanged.

tests/gem_pose.sx now pins the new envelope (locked rest endpoints, the
anticipation dip, the overshoot, the monotonic post-peak collapse) with its
expected snapshot updated.

goldens/p18_pop.png: new mid-clear pop golden at the overshoot peak
(M3TE_FX=3 M3TE_ANIM_TIME=0.21). goldens/p6_fx_match.png refreshed for the
new pop shape (same scene at its documented 0.22). README documents P18.1.

README.md

@@ -384,6 +384,40 @@ env SIMCTL_CHILD_M3TE_FX=11 SIMCTL_CHILD_M3TE_ANIM_TIME=3.0 \
 no column has landed yet, so neither carries squash. The change is render-only — no
 `board.sx` model change, and a resting board is untouched.
 
+### Organic combine — anticipation pop on clear (P18.1)
+
+Matched gems no longer just pop-then-shrink flatly: `clear_pop_scale` now shapes the
+clear as a candy pop in three beats over its local `0..1` — a tiny anticipation
+squash dip (a "gather" ~8 % below rest), a snappy overshoot up to ~1.40× via P15.1's
+`ease_out_back`, then an accelerating collapse to nothing (`ease_in_quad`). The
+endpoints stay LOCKED — `t==0 → 1.0` (rest) and `t==1 → 0.0` (gone) — so the seam to
+the model board is clean and `M3TE_ANIM_TIME=0` still reproduces the resting board;
+the soft particle burst / `+points` popup (`board_fx.sx`) compose on top.
+`tests/gem_pose.sx` pins the new envelope (locked rest endpoints, the anticipation
+dip below rest, the overshoot above 1, and the strictly monotonic post-peak
+collapse). `CLEAR_ANIM_DUR` (0.14 s) is unchanged, so the per-round cascade-cue
+timing snapshots (`tests/cascade_rounds.sx` / `cascade_cue.sx`) don't churn.
+
+The pop peaks at clear-phase local `t ≈ 0.37`; for `M3TE_FX=3` (the seed-1337
+vertical red 3-match) the clear window is `[0.16, 0.30)` s, so `0.21` catches the
+matched gems at their fullest overshoot, composed with the burst and "+30" popup:
+
+```bash
+# Anticipation/overshoot candy pop at its peak (composed with the burst):
+#   goldens/p18_pop.png
+env SIMCTL_CHILD_M3TE_FX=3 SIMCTL_CHILD_M3TE_ANIM_TIME=0.21 \
+    xcrun simctl launch --terminate-running-process booted co.swipelab.m3te
+# Same match at exact rest (t=0) — board sits at its resting pose, no pop:
+env SIMCTL_CHILD_M3TE_FX=3 SIMCTL_CHILD_M3TE_ANIM_TIME=0 \
+    xcrun simctl launch --terminate-running-process booted co.swipelab.m3te
+```
+
+`goldens/p6_fx_match.png` (the P11.1 burst+popup reference, pinned at `0.22`) was
+refreshed for the new pop shape. The per-gem STAGGER of the explosions is the next
+step (P18.2); this step is the per-gem pop SHAPE only. The change is render-only —
+no `board.sx` model change, and normal play is byte-identical apart from the clear's
+pop curve.
+
 ### FPS counter — dev overlay (P20.1)
 
 A small FPS readout for gauging frame cost while tuning the animations. It is a

gem_anim.sx

@@ -76,22 +76,39 @@ land_squash :: (tl: f32) -> f32 {
 }
 
 // --- Clear pop ---------------------------------------------------------------
-// The matched-gem clear: a snappy outward pop then a collapse to nothing over its
+// The matched-gem clear, shaped as a candy pop in three beats over its local
-// local 0..1, so the clear reads as a satisfying candy pop rather than a plain
+// 0..1: a tiny anticipation squash dip (a "gather" just below rest), a snappy
-// shrink. A fast rise to a bigger overshoot makes the snap read; the soft
+// overshoot up to the peak via P15.1's ease_out_back, then an accelerating
-// particle burst / score popup (board_fx.sx) compose on top.
+// collapse to nothing. Endpoints are LOCKED — t==0 -> 1.0 (rest) and t==1 -> 0.0
-CLEAR_POP_A    :f32: 0.34;   // overshoot height above resting scale
+// (gone) — so the seam to the model board stays clean; the soft particle burst /
-CLEAR_POP_RISE :f32: 0.18;   // fraction of the window spent rising to the peak
+// score popup (board_fx.sx) compose on top.
+CLEAR_DIP_T    :f32: 0.16;   // fraction of the window spent on the anticipation dip
+CLEAR_DIP_A    :f32: 0.08;   // how far the gem compresses below rest before popping
+CLEAR_POP_RISE :f32: 0.52;   // window fraction at which the overshoot peak is reached
+CLEAR_POP_A    :f32: 0.36;   // overshoot height above resting scale
 
 clear_pop_scale :: (t: f32) -> f32 {
     if t <= 0.0 { return 1.0; }
     if t >= 1.0 { return 0.0; }
-    if t < CLEAR_POP_RISE {
+    if t < CLEAR_DIP_T {
-        return 1.0 + CLEAR_POP_A * (t / CLEAR_POP_RISE);
+        // Anticipation gather: sin(PI*u) is 0 at both ends, so t==0 stays exactly
+        // at rest and the dip hands off to the rise at rest — a brief squash, not
+        // a step.
+        u := t / CLEAR_DIP_T;
+        return 1.0 - CLEAR_DIP_A * sin(PI * u);
     }
+    if t < CLEAR_POP_RISE {
+        // Snap up to the peak: ease_out_back rises from rest, shoots a touch past
+        // 1+A, then eases back to exactly 1+A at the seam (its locked f(1)=1), so
+        // the maximum is a single clean overshoot with no second reversal.
+        u := (t - CLEAR_DIP_T) / (CLEAR_POP_RISE - CLEAR_DIP_T);
+        return 1.0 + CLEAR_POP_A * ease_out_back(u);
+    }
+    // Collapse to nothing: accelerate the shrink from the peak so the gem vanishes
+    // as the burst takes over. ease_in_quad pins the seam at the peak and t==1 at 0.
     peak := 1.0 + CLEAR_POP_A;
     u := (t - CLEAR_POP_RISE) / (1.0 - CLEAR_POP_RISE);
-    peak * (1.0 - u * u)
+    peak * (1.0 - ease_in_quad(u))
 }
 
 // Live per-gem animation state, heap-allocated (like BoardAnim/BoardFx) so it

tests/expected/gem_pose.stdout

@@ -7,7 +7,7 @@ select_start_rest true select_end_rest true select_mid_pops true
 == land squash envelope ==
 land_start_rest true land_end_rest true land_mid_wobbles true
 == clear pop envelope ==
-clear_start_full true clear_end_gone true clear_overshoots true
+clear_start_full true clear_end_gone true clear_dips true clear_overshoots true clear_collapses true
 == gem motion land bookkeeping ==
 motion_init true motion_no_land true motion_fresh_land true
 == gem motion restart resets landings ==

tests/gem_pose.sx

@@ -71,15 +71,30 @@ main :: () -> s32 {
     if !l_end { fails += 1; }
     if !l_mid { fails += 1; }
 
-    // 5. Clear pop: full at t=0, gone at t=1, overshoots above 1 in between.
+    // 5. Clear pop: full at t=0, gone at t=1, a tiny anticipation dip BELOW rest in
+    //    the gather window, an overshoot above 1 mid-window, then — past the peak —
+    //    a strictly monotonic collapse to nothing (a single clean pop, no second
+    //    reversal). The locked t=0/t=1 endpoints keep the seam to the model board.
     print("== clear pop envelope ==\n");
     c_start := approx(clear_pop_scale(0.0), 1.0);
     c_end := approx(clear_pop_scale(1.0), 0.0);
+    c_dip := clear_pop_scale(CLEAR_DIP_T * 0.5) < 1.0;
     c_peak := clear_pop_scale(0.30) > 1.1;
-    print("clear_start_full {} clear_end_gone {} clear_overshoots {}\n", c_start, c_end, c_peak);
+    c_collapse := true;
+    pc := clear_pop_scale(CLEAR_POP_RISE);
+    for 1..21: (i) {
+        tt := CLEAR_POP_RISE + (1.0 - CLEAR_POP_RISE) * cast(f32) i / 20.0;
+        vv := clear_pop_scale(tt);
+        if vv > pc + 0.000001 { c_collapse = false; }
+        pc = vv;
+    }
+    print("clear_start_full {} clear_end_gone {} clear_dips {} clear_overshoots {} clear_collapses {}\n",
+          c_start, c_end, c_dip, c_peak, c_collapse);
     if !c_start { fails += 1; }
     if !c_end { fails += 1; }
+    if !c_dip { fails += 1; }
     if !c_peak { fails += 1; }
+    if !c_collapse { fails += 1; }
 
     // 6. GemMotion land bookkeeping: fresh state is unpinned at t=0, a never-
     //    landed cell rests, and a freshly-stamped land reads age 0.