A smoother stroll through the village

Can one step at a time feel like a continuous stroll? This increment paces received route points, buffers the first continuing step for 0.22 seconds, brakes before known bends and anticipates facing while preserving the exact simulator-approved polyline. The final approach completes even float-precision residuals at distant coordinates. Body cadence uses actual travel and local leg geometry with a soft return to standing. Python adds only a public current-scene route_pending boolean; future home path coordinates no longer travel in the presentation response. Simulation actions, provider budgets and both disabled Needle gates are unchanged. Iteration 00-before preserves sampled walk capture 20261016T062514109809-25b3ea3c0a47 from source-05d9bbfdb69e; iteration 01-cadence-and-stride preserves 20261016T062828549146-7ecf7afccd19 from source-b311771bca98. Iteration 02-precision-and-turns uses final source-e2cafb35c149 with actual mixed-scale route-studio capture 20261016T063510099188-a30379068660; iteration 03-native-build records rebuilt normal-game appearance in 20261016T064023199159-95e81f2c9851. Every image and sidecar was inspected and individually privacy approved. Earlier pose views share a camera; the final route studio and native scene have different framing. Measured 16 October 2026 on source-e2cafb35c149: 35/35 deterministic locomotion checks at 06:34:59 UTC, covering 30/60/120 Hz, late/repeated snapshots, bounded deltas, reset, corner anticipation, exact routes and final positions; 140/140 animated checks at 06:35:30–06:36:02 UTC in the advancing isolated Editor across all 4 original designs at adult/child scale and an unrigged fallback. Near-stop frames below 0.2 m/s fell from 15.37% to 0% for the 60 Hz paced fixture versus the same-build unpaced control, over 540 measured frames from seconds 2–11; this is not an old-build or live-network measurement. 713/713 Python regression tests passed in 49.76 s on the same unchanged source fingerprint. Private immutable reports retain measurement scope and failures. The preliminary 32/35 timing result used an unsuitable literal-stop comparison; the final test preserves stationary counts and measures near-stops and speed variation. The first animated pass exposed genuine distant-coordinate arrival stalls; failed reports remain retained and the original strict assertions now pass. An early preliminary stale source-label invocation has a private correction note and is not used as exact-source evidence. Native build unity-source-e2cafb35c149-20261016063758 succeeded at 06:38:27 UTC in 28.624 s, zero errors and one expected Pipeline-disabled warning; native screenshot captured 06:40:23 UTC on Metal/URP. A separate photo client observed ordinary running life and was closed without pausing it. The local 12-second motion preview retains all 288 source frames, actual transforms, compiled modules and a reviewed media receipt; it is a continuous deterministic studio sequence, without cuts, loops, audio, simulation writes or inference. Stills alone do not establish movement or sustained performance. The original running player and service were left running: relaunch the rebuilt game and restart the local runtime to load the full update, including the new continuing-route boolean. The old service uses the paced fallback without that startup-buffer hint; complete new-contract live acceptance there remains pending. Independent planted feet/IK, slope adaptation, dynamic avoidance and full roadmap 2.2 acceptance remain open. Next useful check: sustained native trips through uneven paths, thresholds and crowded narrow spaces. See docs/resident-movement.md. Ownership was split between movement implementation/integration, a read-only audit and movement-check contributor, the Python presentation-signal contributor and the isolated studio contributor. Each contributing agent acknowledged reading the documentation standard and overview workflow. The app inventory showed no other active project chats, and idle chats were not awakened.