restructure: proper project layout, README, kill training

- cuda/ — main LBM kernel (khra_gixx_1024_v5.cu)
- navigator/ — lattice_observer, golden_weave, bridges, mock daemon
- scripts/ — compile, start, launch, setup (paths updated)
- docs/ — system manual
- archive/ — everything else (old kernels, inquiries, experiments)
- README.md — full setup guide: requirements, quick start, use your own LLM
- removed training/ entirely (broken LoRA scripts + datasets)
- .gitignore: exclude build/ logs/ training/ *.jsonl

archive/misc/compile_v3.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+# compile_v3.sh — Build khra_gixx_1024_v3 (bidirectional + NVML + CHECKPOINT)
+export PATH=/usr/local/cuda-12.6/bin:$PATH
+export LD_LIBRARY_PATH=/usr/local/cuda-12.6/lib64:$LD_LIBRARY_PATH
+
+cd /mnt/d/fractal-brain/beast-build
+
+echo "=== COMPILING khra_gixx_1024_v3 ==="
+echo "  PUB telemetry on 5556 | SUB commands on 5557 | NVML hardware"
+echo "  CHECKPOINT: save_state/load_state/set_autosave (default 100k cycles)"
+echo "nvcc: $(nvcc --version 2>&1 | grep release)"
+echo "Target: sm_89 (RTX 4090)"
+echo ""
+
+nvcc -o khra_gixx_1024_v3 khra_gixx_1024_v3.cu \
+    -lzmq -lnvidia-ml \
+    -O3 -arch=sm_89 \
+    2>&1
+
+if [ $? -eq 0 ]; then
+    echo ""
+    echo "=== BUILD SUCCESS ==="
+    ls -la khra_gixx_1024_v3
+    echo ""
+    echo "To run (fresh start):"
+    echo "  Kill v2:          kill \$(pgrep -f khra_gixx_1024_v2)"
+    echo "  Start v3:         ./khra_gixx_1024_v3"
+    echo ""
+    echo "To resume from checkpoint:"
+    echo "  Start v3, then:   send {\"cmd\":\"load_state\",\"path\":\"/path/to/ckpt.bin\"}"
+    echo ""
+    echo "Manual save:        send {\"cmd\":\"save_state\"}"
+    echo "Change autosave:    send {\"cmd\":\"set_autosave\",\"interval\":50000}"
+else
+    echo ""
+    echo "=== BUILD FAILED ==="
+    exit 1
+fi