docs: add fluff report

FLUFF_REPORT.md

@@ -0,0 +1,65 @@
+# Codebase Fluff, Placeholder, and Simulation Report
+
+This report lists in hyper-detail every instance across all codebase files (excluding documentation like `ReasonBorn.md`) where there is fluff, missing logic, placeholders, mock/simulation tricks, or introduction comments that indicate incomplete or faked implementations.
+
+This is based on a full, exhaustive scan of the repository.
+
+## 1. Mocks, Simulations, and Faked Implementations
+
+These are instances where the code explicitly simulates or fakes functionality instead of implementing it fully.
+
+### `src/reasonborn/deployment/quantization.py`
+- **Line 109:** Docstring states `Inserts fake-quantization modules that simulate quantization` during Quantization-Aware Training (QAT).
+
+### Test Suite Mocks (Legitimate but technically simulations/mocks)
+- **`tests/test_moe_routing.py`**
+  - **Line 5:** `class MockConfig:` - Mocking the configuration object for testing.
+  - **Line 12:** `moe_layer = SparseMoE(MockConfig())` - Instantiating the layer with the mock.
+- **`tests/test_nested_cot.py`**
+  - **Line 3:** `class MockModel:` - Mocking a full model for nested Chain-of-Thought decomposition tests.
+  - **Line 16:** `"""Test nested CoT tree decomposition without mocking."""` - A comment referencing mocking (even though it claims not to use it here).
+  - **Line 17:** `engine = NestedCoTEngine(MockModel(), max_depth=3)` - Uses the mock model.
+- **`tests/test_ewc_retention.py`**
+  - **Line 4:** `class MockModel(torch.nn.Module):` - Mocking a PyTorch model for Elastic Weight Consolidation tests.
+  - **Line 10:** `model = MockModel()` - Instantiating the mock.
+  - **Line 12:** `class MockConfig:` - Mock configuration.
+  - **Line 22:** `# Set test fisher diagonal to simulated importance weights` - Comment indicating importance weights are simulated for the test.
+- **`tests/test_dp_accounting.py`**
+  - **Line 7:** `# Simulate 100 training steps` - Simulating differential privacy accounting steps.
+- **`tests/test_hybrid_attention.py`**
+  - **Line 5:** `class MockConfig:` - Mocking the configuration object.
+
+## 2. Placeholders and Missing Logic
+
+These are instances where the implementation is explicitly deferred, missing, or indicated by placeholder syntax.
+
+### `src/reasonborn/deployment/quantization.py`
+- **Line 97:** `# Implement actual ROCm quantization kernel calibration` - A comment acting as a TODO/placeholder for missing ROCm kernel calibration.
+- **Line 103:** `# Implementation would use AMD ROCm-specific calibration routines` - A comment explaining what the missing implementation would look like.
+
+### `src/reasonborn/learning/curriculum.py`
+- **Line 115:** `                ...` - An ellipsis placeholder used inside a docstring code example to indicate omitted logic.
+
+## 3. Dummy Variables/Tricks
+
+Variables explicitly named or created as dummies to trick functions into working, usually in test scenarios.
+
+### Test Suite Dummy Variables
+- **`tests/test_moe_routing.py`**
+  - **Line 13:** `dummy_input = torch.randn(2, 64, 256)` - Creating a dummy random tensor.
+  - **Line 15:** `output, aux_loss = moe_layer(dummy_input)` - Passing the dummy variable.
+  - **Line 17:** `assert output.shape == dummy_input.shape` - Asserting against the dummy variable.
+- **`tests/test_hybrid_attention.py`**
+  - **Line 12:** `dummy_input = torch.randn(2, 1024, 128)` - Creating a dummy random tensor.
+  - **Line 13:** `output = layer(dummy_input)` - Passing the dummy variable.
+  - **Line 15:** `assert output.shape == dummy_input.shape` - Asserting against the dummy variable.
+
+## 4. Legitimate Terms Resembling Fluff (False Positives Explained)
+
+During the exhaustive search, several items were flagged but are actually fully legitimate structural elements and not "fluff":
+
+- **Pass statements in exception handling:** Lines such as `pass` inside `except Exception:` blocks (e.g., `src/reasonborn/reasoning/verification/consistency.py` Line 135) are legitimate control flow.
+- **Forward/Backward pass:** Comments mentioning "forward pass" or "backward pass" refer to legitimate neural network operations.
+- **Pass/Passed statuses:** Dictionary keys like `"passed": True` inside the reasoning and verification modules represent legitimate boolean statuses, not fluff.
+- **Safety Filter bypass:** `src/reasonborn/control/safety_filter.py` Line 23 contains the regex `r'\b(how to|instructions for)\s+(hack|break into|bypass)'`. This is a legitimate security rule string, not a trick in the codebase.
+- **Pass unmodified:** `src/reasonborn/control/safety_filter.py` Line 305 contains `# FACTUAL claims pass through unmodified`, which is a legitimate architectural comment.