Drill: visible dust; pulses push loose material back and open a channel (t92)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs/DESIGN.md

@@ -542,6 +542,12 @@ above, including:
   flow, pockets drain), with a little dust. All damping is counted per grain travelled
   (attenuation is per distance in real physics), retuned so rock bores feel the same as before.
 
+- **Pulses stream loose grains back toward the bot (t92 follow-up, 2026-10-28).** Owner
+  playtesting found that surface bouncing did not open a channel. Keep shaking the travelled
+  ray path while the pulse is in flight, bias short hops back toward the bot, and keep sideways
+  and upward escapes when blocked. Grains fall normally after each hop. Dust must show as a few
+  soft puffs where grains move; the previous rubble-coloured debris squares were not visible.
+
 ## Ideas not built yet
 
 - **Concrete**, as a stronger wall material from rubble.

docs/PROJECT_MAP.md

@@ -327,7 +327,7 @@ Tool helpers: reach, brushes, and plain digging.
 - functions: setBuildRange, inBuildRange, inReach, forBrush, drillReach, inDrillReach, toolDig
 - variables: BUILD_MAX, buildBox
 
-### src/sim/drill.js (187 lines)
+### src/sim/drill.js (202 lines)
 The resonance drill: tuned pulses that shatter one material.
 - functions: drillOrigin, pulseRay, builtCell, toolPulse, crumbleAround, crumbly, shatterCell, pulseTick, pingRadius, resonancePing
 - variables: PULSE_W, PULSE_EVERY, PULSE_SPEED, PULSE_K, drillPulses, lastPulseTick, drillWidth, THIN_BONUS, CRUMBLE_MAX, THIN_BONUS_OFF, PING_VEINS
@@ -482,7 +482,7 @@ Scale colour without changing the smooth-look class.
 Machine overlays: fans, engines, silo and the rest, drawn on top of the world.
 - functions: cellRect, drawHazardRims, drawMachines, drawProcBlocks, drawSilo, drawSiloPixel, drawSiloWindow, drawRackBlocks
 
-### src/render/overlays.js (370 lines)
+### src/render/overlays.js (384 lines)
 Other overlays: floating text, rain and snow, flow arrows, filter tags, molds, the bot, particles.
 - functions: spawnFloater, drawRain, drawFlowArrows, drawJoinMark, drawJoinMarks, flowLoaded, drawFlowRates, pingImage, drawPing, drawHoveredVein, drawPingRecharge, drawRunOutline, drawFilterTags, drawFloaters, moldLabel, drawMolds, drawBot, spawnDebris, spawnSpark, spawnSuck, drawParticles
 - variables: PING_SPEED, PING_HOLD, PING_FADE, PING_BANDS, PING_GLINTS, GLINT_MS

docs/agent-log.md

@@ -4,6 +4,46 @@ What agents have done, newest first. Add an entry for every piece of work: the d
 note or request it was for, what changed, the files touched, the tests run, and anything left
 open or not verified. Keep each entry short; the git history has the details.
 
+
+## 2026-10-28: t92 follow-up, visible dust and a botward channel through rubble (Codex)
+
+- Traced shake events through `drainEvents` to `spawnDebris` and `drawParticles`. Events were
+  handled above the world, but rendered only two opaque, rubble-coloured 0.7-grain squares.
+  Added a dust flag to the existing debris event/particle pipeline: one pale radial puff,
+  gently rising/expanding and fading over 36 frames. Retained particles setting and particle
+  limit; at most four shake dust events per tick. Normal shatter debris/sparks unchanged.
+- Kicks previously visited only the advancing 2.2-grain wave front, with random sideways/up
+  directions. Now each in-flight ray keeps shaking its travelled path, with local energy
+  attenuated through powder/liquid; stops at solid. Tries the reverse ray first, then lift and
+  side escapes. Retains seeded rnd(), one-grain move/enter hops, occupied/part/stamp guards,
+  sealed corners and three moves/ray/tick; gravity settles grains normally. No damping retune.
+- Added steel-medium 35-grain-deep pile/rock regression and gravity-only control. At 480 ticks:
+  33/35 centre-line cells empty versus 9/35; rubble centre x 72 -> 56.168 versus 70.68;
+  all 875 rubble grains retained. The original pulse code gives only 10/35 empty cells and
+  centre x 68.341, failing channel, transport, dust and travelled-path checks. An isolated
+  advanced-front ray checks interior botward movement; existing movement/dust caps and guards,
+  attenuation, repeatability and heap-fluidising tests retained. New browser check exercises
+  event -> visible soft pixels, expiry and particles-off.
+- Stone depth/opened cells against original t92 baseline: narrow 22/181 -> 24/195,
+  medium 13/186 -> 13/198, wide 6/134 -> 6/133; all within 10%, including depth.
+  Small 480-tick pile benchmark median, five measured samples after warmup: 24.18 -> 33.88 ms
+  (about 0.02 ms/tick extra). This is a small fixture, not a whole-game performance profile.
+- Full `python3 test/run_all.py --browser` exited 0, ALL PASSED: 32 headless suites and
+  gameplay 361, smooth 28, console 18, caps 7, GPU 15, goldens 28, light 71, Firefox 2.
+  Final pile fixture was moved into the real steel reach and its conservation assertion added
+  after that run; the final `python3 test/run_all.py` passes all 32 suites, drill 66 checks.
+  Inspected final `shots/t92b-channel.png` and `shots/t92b-channel-smooth.png`: lowered
+  channel, basalt notch and soft dust over the streaming rubble. First recapture timed out
+  starting the world; host-GPU recaptures passed. Caps confirmed NVIDIA Blackwell; no claim
+  that the timeout was caused by unavailable hardware. Screenshot setup: `/tmp/t92b-channel.js`,
+  `tools/shot.py --setup /tmp/t92b-channel.js --zoom 8 --no-light --freeze` (add smooth look).
+  Evidence: `/tmp/t92b-red-final.log`, `/tmp/t92b-drill.log`, `/tmp/t92b-all-tests.log`,
+  `/tmp/t92b-headless-final.log`. No page errors on final captures.
+- Files: `src/sim/drill.js`, `src/render/overlays.js`, `src/ui/events.js`, `test/t_drill.js`,
+  `test/t_gpu.py`, generated `docs/PROJECT_MAP.md`, owner decision in `docs/DESIGN.md`,
+  t92 tracking and this log. Both feedback items ticked; t92 moved to Review. No commit.
+  Owner playtesting and performance in a large active world remain unverified.
+
 ## 2026-10-28: t92 drill pulses pass through rubble and fluidise it (Codex)
 
 - Converted every pulse loss to a per-grain factor, applied with `factor ** step` outside