2.3

Adjusted Signal

2.3

Raw Pattern Score

100%

Time Factor

2026-10-28

Last Push

15.6K

Stars

C++

Language

407.8K

Lines of Code

820

Files

892

Pattern Hits

2026-10-31

Scan Date

0.01

HC Hit Rate

Risk Review · Payload v1

Use each vector to guide manual verification of the underlying findings.

Payload schema: 1.0.0 · Engine: 2.0.0 · Scoring: heuristic-density-v1 · Scan date: 2026-10-31.

Cognitive Ownership Risk 10.6/100 · LOW

Can maintainers explain intent, assumptions, and trade-offs? 878 finding(s).

Hallucinated Dependency & Supply Chain Risk 0.0/100 · LOW

Are imports declared and package/API names verified? 3 finding(s).

Agent Over-Engineering Ratio 0.5/100 · LOW

Are wrappers, guards, and repeated structures necessary? 22 finding(s).

PR Review Velocity Impact 10.8/100 · LOW

Which changes need focused human review? 892 finding(s).

What These Metrics Mean

Adjusted Signal: Legacy pattern score normalized per 1,000 scanned lines and multiplied by the repository-level time factor. It is not an estimate of AI authorship.
Raw Pattern Score: Sum of severity- and context-weighted pattern findings before applying the time factor.
Time Factor: A coarse multiplier based on repository last-push metadata. It does not identify when individual code was written.
Pattern Hits: Count of configured pattern findings. Review each location and surrounding code before acting.
HC Hit Rate: High- and critical-severity findings per scanned file. Severity is a configured detector weight, not proof of a vulnerability or authorship.
Lines / Files: Scan coverage denominators used by the legacy normalized score.

Pattern-Signal History

This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.

Severity Breakdown

Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.

CRITICAL 0HIGH 5MEDIUM 9LOW 878

Directory Score Breakdown

This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.

Pattern Findings

The scanner identified 892 distinct pattern matches across 16 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.

Over-Commented Block789 hits · 763 pts

Severity	File	Line	Snippet	Context
LOW	ci/run.sh	1	#/bin/bash	COMMENT
LOW	ci/run.sh	241	gg_printf '```\n'	COMMENT
LOW	ci/run.sh	261	#	COMMENT
LOW	include/ggml-cpp.h	1	#pragma once	COMMENT
LOW	include/ggml-sycl.h	1	//	COMMENT
LOW	include/ggml-backend.h	1	#pragma once	COMMENT
LOW	include/ggml-backend.h	321	// if the user returns false the scheduler simply copies the entire weight	COMMENT
LOW	include/ggml-backend.h	401	GGML_API const char * ggml_backend_meta_split_axis_name(enum ggml_backend_meta_split_axis split_axis);	COMMENT
LOW	include/ggml-cuda.h	1	#pragma once	COMMENT
LOW	include/ggml-opt.h	1	// This file contains functionality for training models using GGML.	COMMENT
LOW	include/ggml-opt.h	181	struct ggml_tensor * inputs,	COMMENT
LOW	include/ggml-metal.h	1	// Note: this description is outdated	COMMENT
LOW	include/ggml-metal.h	21		COMMENT
LOW	include/gguf.h	1	// This file contains functionality related to "GGUF" files, the binary file format used by ggml.	COMMENT
LOW	include/gguf.h	21	// 1. The size of the tensor in the dimension (int64_t).	COMMENT
LOW	include/gguf.h	161		COMMENT
LOW	include/gguf.h	181	// FILE * f = fopen(fname, "ab");	COMMENT
LOW	include/ggml-rpc.h	1	#pragma once	COMMENT
LOW	include/ggml.h	1	#pragma once	COMMENT
LOW	include/ggml.h	21	//	COMMENT
LOW	include/ggml.h	41	// struct ggml_context * ctx = ggml_init(params);	COMMENT
LOW	include/ggml.h	61	// struct ggml_cgraph * gf = ggml_new_graph(ctx);	COMMENT
LOW	include/ggml.h	81	// and after defining the computation graph, call the ggml_used_mem() function to find out how much memory was	COMMENT
LOW	include/ggml.h	101	// - ggml_conv_1d_2s()	COMMENT
LOW	include/ggml.h	121	// assert(c->src[0] == a);	COMMENT
LOW	include/ggml.h	141	// (float ) ((char ) a->data + ya->nb[1] + x*a->nb[0]) = x + y;	COMMENT
LOW	include/ggml.h	161	//	COMMENT
LOW	include/ggml.h	181	# define GGML_API __declspec(dllimport) extern	COMMENT
LOW	include/ggml.h	201	#elif defined(__MINGW32__) && !defined(__clang__)	COMMENT
LOW	include/ggml.h	221		COMMENT
LOW	include/ggml.h	241	#define GGML_MEM_ALIGN 8	COMMENT
LOW	include/ggml.h	261	__host__ __device__ constexpr inline void ggml_unused_vars_impl(Args&&...) noexcept {}	COMMENT
LOW	include/ggml.h	281	#elif defined(_MSC_VER)	COMMENT
LOW	include/ggml.h	421	// GGML_TYPE_Q4_0_4_4 = 31, support has been removed from gguf files	COMMENT
LOW	include/ggml.h	1441	struct ggml_tensor * a,	COMMENT
LOW	include/ggml.h	1461	// ranks in decreasing order:	COMMENT
LOW	include/ggml.h	1881	int n_dims,	COMMENT
LOW	include/ggml.h	1921	float attn_factor,	COMMENT
LOW	include/ggml.h	1941	// given sections = [t=4, y=2, x=2, 0]	COMMENT
LOW	include/ggml.h	2201	int d1, // dilation height	COMMENT
LOW	include/ggml.h	2681	// DSA lightning indexer	COMMENT
LOW	include/ggml.h	2701	// using a richer representation through streams.	COMMENT
LOW	include/ggml.h	2741	GGML_API struct ggml_tensor * ggml_dsv4_hc_post(	COMMENT
LOW	include/ggml.h	2881	// this is useful for creating graphs that have constant topology but compute different things based on the input	COMMENT
LOW	include/ggml.h	2961	GGML_API void ggml_log_set(ggml_log_callback log_callback, void * user_data);	COMMENT
LOW	include/ggml.h	3001	# define GGML_RESTRICT __restrict	COMMENT
LOW	tests/test-conv2d.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-timestep_embedding.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-conv1d-dw-c1.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-conv1d-dw-c2.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-pad-reflect-1d.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-quantize-fns.cpp	1	// Unit tests for quantization specific functions - quantize, dequantize and dot product	COMMENT
LOW	tests/test-roll.cpp	1	#include <ggml.h>	COMMENT
LOW	tests/test-opt.cpp	1	// TODO refactor	COMMENT
LOW	tests/test-conv-transpose-1d.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-quantize-perf.cpp	1	// Benchmark quantization specific functions on synthetic data	COMMENT
LOW	tests/test-arange.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-customop.c	1	#include "ggml.h"	COMMENT
LOW	tests/test-conv1d.cpp	1	#include "ggml.h"	COMMENT
LOW	tests/test-backend-ops.cpp	1	// This file defines tests for various GGML ops and backends.	COMMENT
729 more matches not shown…

Structural Annotation Overuse32 hits · 57 pts

Severity	File	Line	Snippet	Context
LOW⚡	tests/test-backend-ops.cpp	2071	// Step 1: create input tensors that don't depend on any other tensors:	COMMENT
LOW⚡	tests/test-backend-ops.cpp	2078	// Step 2: use the op that you want to test in the GGML compute graph.	COMMENT
LOW⚡	tests/test-backend-ops.cpp	2082	// Step 3: return the output tensor.	COMMENT
LOW	src/ggml-et/et-kernels/src/norm_f32.c	84	// Step 1: sum for mean	COMMENT
LOW	src/ggml-et/et-kernels/src/norm_f32.c	110	// Step 2: compute (x - mean) → dst, accumulate variance	COMMENT
LOW	src/ggml-et/et-kernels/src/norm_f32.c	143	// Step 3: apply scale to centered values in dst	COMMENT
LOW	src/ggml-et/et-kernels/src/rms_norm_f32.c	100	// Step 1: Compute sum of squares using 8-wide vectors	COMMENT
LOW⚡	src/ggml-et/et-kernels/src/rms_norm_f32.c	124	// Step 2: scale = rsqrt(mean + eps)	COMMENT
LOW⚡	src/ggml-et/et-kernels/src/rms_norm_f32.c	132	// Step 3: Apply scaling: broadcast scale once, reuse across loop	COMMENT
LOW	src/ggml-et/et-kernels/src/rwkv_wkv7_f32.c	177	// Step 1: sa = dot(a, state_prev[i])	COMMENT
LOW	src/ggml-et/et-kernels/src/rwkv_wkv7_f32.c	196	// Step 2: state update + result accumulation	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	318	// Step 1: Apply decay gate to state	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	323	// Step 2: delta = (v - S @ k) * beta	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	330	// Step 3: state += outer(delta, k)	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	336	// Step 4: attn_out = S @ q * scale	COMMENT
LOW⚡	src/ggml-metal/CMakeLists.txt	118	# Step 1: concatenate shared headers + this kernel source	COMMENT
LOW⚡	src/ggml-metal/CMakeLists.txt	120	# Step 2: remove internal #include and #pragma once	COMMENT
LOW⚡	src/ggml-metal/CMakeLists.txt	122	# Step 3: inline ggml-common.h (replacing __embed_ggml-common.h__ sentinel)	COMMENT
LOW⚡	src/ggml-metal/CMakeLists.txt	124	# Step 4: inline ggml-metal-impl.h	COMMENT
LOW⚡	src/ggml-metal/CMakeLists.txt	126	# Step 5: emit an asm chunk with kind-specific start/end symbols	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	2853	// Step 5: multiply by attn_factor	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	2870	// Step 6: repeat	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3134	// Step 1: Prepare trans tensors for F16 type conversion to F32 if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3159	// Step 2: Prepare head tensors for tail splitting if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3188	// Step 3: Execute RotaryPositionEmbedding	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3235	// Step 4: Copy unrotated tail portion from source to destination	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3264	// Step 5: Cast back to F16 if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3950	// Step 1: cast the src0 (Query) to fp16 if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3973	// Step 2: create the acl tensors for src1 (Key), src2 (Value),	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	4017	// Step 3: create the PSEShift tensor if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	4086	// Step 4: set the inputs for FusedInferAttention.	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	4147	// Step 6: post-processing — slice padded output and/or cast to f32	COMMENT

Verbosity Indicators28 hits · 47 pts

Severity	File	Line	Snippet	Context
LOW⚡	tests/test-backend-ops.cpp	2071	// Step 1: create input tensors that don't depend on any other tensors:	COMMENT
LOW⚡	tests/test-backend-ops.cpp	2078	// Step 2: use the op that you want to test in the GGML compute graph.	COMMENT
LOW⚡	tests/test-backend-ops.cpp	2082	// Step 3: return the output tensor.	COMMENT
LOW	src/ggml-et/et-kernels/src/norm_f32.c	84	// Step 1: sum for mean	COMMENT
LOW	src/ggml-et/et-kernels/src/norm_f32.c	110	// Step 2: compute (x - mean) → dst, accumulate variance	COMMENT
LOW	src/ggml-et/et-kernels/src/norm_f32.c	143	// Step 3: apply scale to centered values in dst	COMMENT
LOW	src/ggml-et/et-kernels/src/rms_norm_f32.c	100	// Step 1: Compute sum of squares using 8-wide vectors	COMMENT
LOW⚡	src/ggml-et/et-kernels/src/rms_norm_f32.c	124	// Step 2: scale = rsqrt(mean + eps)	COMMENT
LOW⚡	src/ggml-et/et-kernels/src/rms_norm_f32.c	132	// Step 3: Apply scaling: broadcast scale once, reuse across loop	COMMENT
LOW	src/ggml-et/et-kernels/src/rwkv_wkv7_f32.c	177	// Step 1: sa = dot(a, state_prev[i])	COMMENT
LOW	src/ggml-et/et-kernels/src/rwkv_wkv7_f32.c	196	// Step 2: state update + result accumulation	COMMENT
LOW	src/ggml-webgpu/ggml-webgpu.cpp	1454	// For set rows specifically, we need to check if src and idx are empty	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	318	// Step 1: Apply decay gate to state	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	323	// Step 2: delta = (v - S @ k) * beta	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	330	// Step 3: state += outer(delta, k)	COMMENT
LOW⚡	src/ggml-openvino/openvino/op/gated_delta_net.cpp	336	// Step 4: attn_out = S @ q * scale	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	2853	// Step 5: multiply by attn_factor	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	2870	// Step 6: repeat	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3134	// Step 1: Prepare trans tensors for F16 type conversion to F32 if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3159	// Step 2: Prepare head tensors for tail splitting if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3188	// Step 3: Execute RotaryPositionEmbedding	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3235	// Step 4: Copy unrotated tail portion from source to destination	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3264	// Step 5: Cast back to F16 if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3950	// Step 1: cast the src0 (Query) to fp16 if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	3973	// Step 2: create the acl tensors for src1 (Key), src2 (Value),	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	4017	// Step 3: create the PSEShift tensor if needed	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	4086	// Step 4: set the inputs for FusedInferAttention.	COMMENT
LOW	src/ggml-cann/aclnn_ops.cpp	4147	// Step 6: post-processing — slice padded output and/or cast to f32	COMMENT

Cross-File Repetition4 hits · 20 pts

Severity	File	Snippet	Context
HIGH	examples/gpt-2/convert-ckpt-to-ggml.py	returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings	STRING
HIGH	examples/gpt-2/convert-h5-to-ggml.py	returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings	STRING
HIGH	examples/gpt-2/convert-cerebras-to-ggml.py	returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings	STRING
HIGH	examples/gpt-j/convert-h5-to-ggml.py	returns list of utf-8 byte and a corresponding list of unicode strings. the reversible bpe codes work on unicode strings	STRING

AI Slop Vocabulary4 hits · 12 pts

Severity	File	Line	Snippet	Context
MEDIUM	examples/mnist/mnist-common.h	20	// The physical batch size determines how many datapoints are processed in parallel, larger values utilize compute bette	COMMENT
MEDIUM	src/ggml-cpu/amx/mmq.cpp	603	// pack again with 128 to fully utilize vector length	COMMENT
MEDIUM	src/ggml-cpu/amx/mmq.cpp	645	// pack again with 128 to fully utilize vector length	COMMENT
MEDIUM	src/ggml-cpu/amx/mmq.cpp	743	// pack again with 128 to fully utilize vector length	COMMENT

Hyper-Verbose Identifiers10 hits · 10 pts

Severity	File	Line	Snippet	Context
LOW	include/ggml-backend.h	215	typedef ggml_backend_buffer_type_t (ggml_backend_split_buffer_type_t)(int main_device, const float tensor_split	CODE
LOW	examples/mnist/server.py	9	class CustomHTTPRequestHandler(http.server.SimpleHTTPRequestHandler):	CODE
LOW	examples/python/ggml/utils.py	172	def __check_shape_consistent_with_type(tensor: TensorLike):	CODE
LOW⚡	src/ggml-virtgpu/regenerate_remoting.py	49	def _check_clang_format_available(self) -> bool:	CODE
LOW⚡	src/ggml-virtgpu/regenerate_remoting.py	53	def _format_file_with_clang_format(self, file_path: Path) -> bool:	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	78	def generate_backend_function_name(self, group_name: str, function_name: str) -> str:	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	89	def generate_frontend_function_name(self, group_name: str, function_name: str) -> str:	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	125	def generate_apir_backend_header(self) -> str:	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	176	def generate_backend_dispatched_header(self) -> str:	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	222	def generate_virtgpu_forward_header(self) -> str:	STRING

Excessive Try-Catch Wrapping6 hits · 8 pts

Severity	File	Line	Snippet	Context
MEDIUM	examples/mnist/server.py	35	print(f"Error: {e}")	CODE
LOW	examples/python/example_test_all_quants.py	65	except Exception as e:	CODE
MEDIUM	examples/python/example_test_all_quants.py	66	print(f'Error: {e}')	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	69	except Exception as e:	CODE
LOW	src/ggml-virtgpu/regenerate_remoting.py	327	except Exception as e:	STRING
MEDIUM	src/ggml-virtgpu/regenerate_remoting.py	323	def main():	CODE

Example Usage Blocks4 hits · 6 pts

Severity	File	Line	Snippet	Context
LOW	scripts/sync-whisper-am.sh	5	# Usage:	COMMENT
LOW	scripts/release.sh	11	# Usage:	COMMENT
LOW	scripts/sync-llama-am.sh	5	# Usage:	COMMENT
LOW	src/ggml-sycl/dpct/helper.hpp	3020	/// Example usage:	COMMENT

TODO Padding3 hits · 4 pts

Severity	File	Line	Snippet	Context
LOW	src/ggml.c	4015	// TODO: implement non F32 return	COMMENT
LOW	src/ggml.c	4039	// TODO: implement non F32 return	COMMENT
LOW	src/ggml.c	5218	// TODO: implement antialias for modes other than bilinear	COMMENT

Redundant / Tautological Comments2 hits · 4 pts

Severity	File	Line	Snippet	Context
LOW	scripts/release.sh	77	# Check if we have the latest from master (skip in dry-run)	COMMENT
LOW⚡	src/ggml-virtgpu/regenerate_remoting.py	46	# Check if clang-format is available	COMMENT

Self-Referential Comments1 hit · 3 pts

Severity	File	Line	Snippet	Context
MEDIUM	ci/run.sh	367	# Create a fresh python3 venv and enter it	COMMENT

Unused Imports3 hits · 3 pts

Severity	File	Line	Context
LOW	examples/python/stubs.py	7	CODE
LOW	examples/gpt-2/convert-cerebras-to-ggml.py	9	CODE
LOW	examples/gpt-j/convert-h5-to-ggml.py	23	CODE

Deep Nesting3 hits · 3 pts

Severity	File	Line	Context
LOW	examples/python/ggml/utils.py	118	CODE
LOW	src/ggml-webgpu/wgsl-shaders/embed_wgsl.py	68	CODE
LOW	src/ggml-hexagon/htp/htp-swiv.py	409	CODE

Decorative Section Separators1 hit · 3 pts

Severity	File	Line	Snippet	Context
MEDIUM	src/ggml-et/et-kernels/src/platform.h	347	// ────── ──────────── ──────────────────────────────────────────────────	COMMENT

Synthetic Comment Markers1 hit · 2 pts

Severity	File	Line	Snippet	Context
HIGH	src/ggml-virtgpu/regenerate_remoting.py	3	# Generated by Claude AI	STRING

AI Structural Patterns1 hit · 1 pts

Severity	File	Line	Snippet	Context
LOW	examples/mnist/mnist-train-fc.py	31		CODE

ggml-org/ggml

Open-Source Code Quality, Security and Technical Debt Review