Every transport piece moves material at belt speed and capacity (t105)

docs/agent-log.md

@@ -4,6 +4,22 @@ 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.
 
+## 2026-10-30: t105 transport matches the t45 belt (Codex)
+
+- Shared diagonal cadence, six-grain vertical slope beds, taller cleat pockets and four-lane
+  screw handoffs; raised outlets throw clear of the belt lip. Flat belt and worldgen unchanged.
+  Full before/after table in t105: mixed chain 30.0 → 89.9 grains/s, no spills, 7216/7216 grains
+  accounted for. Diagonal path pace is 31.8 versus 30 cells/s straight (integer cadence).
+- Files: sim/{cells,conveyors,items,machine-tick,machines,pieces,slope-conveyors,transport-status};
+  test/{t_conveyors,t_diag_screws,t_slope_belts,t_threads,t_transport_balance,fixtures/t105-transport};
+  tools/{experiments/transport-balance,scenes/t105-chain}; docs/{DESIGN,TESTING,PROJECT_MAP}, t105.
+- All 44 headless suites pass, including 91 transport checks and 0/1/2/8-worker chain equality;
+  42 new checks fail on original code. Two generated-world hashes match the original revision.
+  Host-GPU t_stalls.py, t_smooth.py, play.py, t_gpu_golden.py and t_screw_heat.py pass.
+- Inspected shots/t105-chain{,-smooth}.png: full pockets and beds, connected joins, clear outlet.
+  Map regenerated; measured transport cost in the task report. Sandbox browser startup timed out.
+  Full browser suite not run; owner playtest remains. No commit.
+
 ## 2026-10-29: new belts stood still until reload; golden test after t45 (Claude)
 
 - Owner bug: in a world started after a long session, belts built after loading never moved