chore(phase0): reproducible compute gate — torch cu130 + GPU smoke test

scripts/check_gpu.py verifies PyTorch cu130 sees the koala Blackwell GPU
(sm_120) and computes — the Phase-0 prerequisite before any autoresearch
experiment. Verified green: torch 2.12.1+cu130, RTX 5070, GPU matmul OK.
Note: koala GPU is shared with the llama-swap LLM stack — run the autoresearch
agent on iguana/berget so the card stays free for train.py.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.gitignore

@@ -28,3 +28,6 @@ go.work.sum
 # Project-specific
 bin/
 *.templ.go
+
+# python venv (autoresearch loop)
+.venv/

requirements.txt

@@ -0,0 +1,6 @@
+# Python deps for the autoresearch loop (train.py + scripts). The Go side
+# (data pipeline, eval harness) is separate. Install torch from the cu130 index:
+#   pip install torch --index-url https://download.pytorch.org/whl/cu130
+#   pip install -r requirements.txt
+# koala = Blackwell sm_120, driver R610; torch 2.12.1+cu130 verified 2026-06-23.
+numpy>=2.0

scripts/check_gpu.py

@@ -0,0 +1,21 @@
+"""Phase-0 compute gate (brain wiki/jepa-fx/facts/autoresearch-integration-phase1):
+PyTorch cu130 must see the koala Blackwell GPU and compute before any experiment.
+
+    python scripts/check_gpu.py   # exits 0 if the GPU is usable, 1 otherwise
+
+Note: koala shares this 12GB card with the llama-swap LLM stack. The autoresearch
+agent should run on iguana/berget models so koala's GPU stays free for train.py.
+"""
+import sys
+import torch
+
+print("torch", torch.__version__)
+if not torch.cuda.is_available():
+    print("CUDA NOT AVAILABLE — gate BLOCKED")
+    sys.exit(1)
+print("device:", torch.cuda.get_device_name(0))
+print("capability: sm_%d%d" % torch.cuda.get_device_capability(0))
+x = torch.randn(2000, 2000, device="cuda")
+(x @ x).sum().item()
+torch.cuda.synchronize()
+print("GPU matmul OK — Phase-0 compute gate GREEN")