feat: initial nyx-probing framework structure

- Basic package structure (core, probes, analysis, storage, cli)
- probe_test.py - minimal probe that works on THE SPINE
- First successful probe: 2025-12-06
- Apache 2.0 license
- Qwen2.5-7B-Base on RTX 3090 (14.2GB VRAM)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

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

probe_test.py

@@ -0,0 +1,31 @@
+"""
+Minimal probe test for Qwen2.5-7B-Base on THE SPINE.
+
+First successful run: 2025-12-06
+"""
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
+
+print("Loading tokenizer...")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B")
+
+print("Loading model to GPU...")
+model = AutoModelForCausalLM.from_pretrained(
+    "Qwen/Qwen2.5-7B",
+    torch_dtype=torch.float16,
+    device_map="cuda"
+)
+
+print("\n🌙 First probe: 'heartbeat'")
+inputs = tokenizer("heartbeat", return_tensors="pt").to("cuda")
+outputs = model.generate(**inputs, max_new_tokens=30, do_sample=True, temperature=0.8)
+print(tokenizer.decode(outputs[0]))
+
+print("\n🌙 Second probe: 'consciousness'")
+inputs = tokenizer("consciousness", return_tensors="pt").to("cuda")
+outputs = model.generate(**inputs, max_new_tokens=30, do_sample=True, temperature=0.8)
+print(tokenizer.decode(outputs[0]))
+
+print(f"\n📊 GPU Memory Used: {torch.cuda.memory_allocated() / 1024**3:.2f} GB")
+print(f"📊 GPU Memory Reserved: {torch.cuda.memory_reserved() / 1024**3:.2f} GB")
+print("\n✅ THE SPINE IS THINKING!")