{
  "_comment": "Manual audit layered on top of the FROZEN rubric (research.md section 4). The rubric label is never altered; this file defines the secondary ai_label_reviewed column. Every override and every kept-as-is judgment call is listed with a reason. Audit performed 2026-08-27 against name/description/topics as served by the API that day; deleted repos judged from name and public record.",
  "overrides": {
    "xai-org/grok-1": {"label": 1, "reason": "open LLM weights release; description 'Grok open release' carries no rubric keyword"},
    "abi/screenshot-to-code": {"label": 1, "reason": "GPT-4V/Claude-driven code generation tool; no rubric keyword in name/desc/topics"},
    "facebookresearch/segment-anything": {"label": 1, "reason": "foundation vision model (SAM)"},
    "chenfei-wu/TaskMatrix": {"label": 1, "reason": "formerly microsoft/visual-chatgpt; LLM+vision agent system"},
    "XingangPan/DragGAN": {"label": 1, "reason": "GAN-based image editing; rubric miss noted ('rag' substring in 'drag' rejected by token rule)"},
    "zai-org/ChatGLM-6B": {"label": 1, "reason": "open LLM"},
    "meta-llama/llama3": {"label": 1, "reason": "open LLM ('llama' not in frozen keyword list)"},
    "meta-llama/llama": {"label": 1, "reason": "open LLM"},
    "meta-llama/codellama": {"label": 1, "reason": "code LLM"},
    "lm-sys/FastChat": {"label": 1, "reason": "LLM serving/eval platform (Vicuna, Chatbot Arena)"},
    "openinterpreter/openinterpreter": {"label": 1, "reason": "LLM coding agent; topics today ('coding-agent','deepseek') miss the exact-match list"},
    "stitionai/devika": {"label": 1, "reason": "'Agentic Software Engineer' — description avoids every rubric keyword"},
    "hpcaitech/Open-Sora": {"label": 1, "reason": "open video-generation model"},
    "cursor/cursor": {"label": 1, "reason": "AI code editor; repo description now empty"},
    "yoheinakajima/babyagi": {"label": 1, "reason": "autonomous LLM agent; repo description now empty"},
    "tloen/alpaca-lora": {"label": 1, "reason": "LLM fine-tuning (LoRA on LLaMA)"},
    "lllyasviel/Fooocus": {"label": 1, "reason": "Stable-Diffusion-XL image generation UI; description 'Focus on prompting and generating'"},
    "KindXiaoming/pykan": {"label": 1, "reason": "Kolmogorov-Arnold networks — neural-network research"},
    "karpathy/llama2.c": {"label": 1, "reason": "LLM inference in C"},
    "HumanAIGC/AnimateAnyone": {"label": 1, "reason": "diffusion-based human animation"},
    "ml-explore/mlx": {"label": 1, "reason": "Apple ML framework"},
    "naklecha/llama3-from-scratch": {"label": 1, "reason": "LLM implementation tutorial"},
    "openai/shap-e": {"label": 1, "reason": "generative 3D model"},
    "togethercomputer/OpenChatKit": {"label": 1, "reason": "open LLM chatbot kit"},
    "WongKinYiu/yolov9": {"label": 1, "reason": "object-detection model"},
    "THU-MIG/yolov10": {"label": 1, "reason": "object-detection model"},
    "instantX-research/InstantID": {"label": 1, "reason": "diffusion-based ID-preserving image generation"},
    "TencentARC/PhotoMaker": {"label": 1, "reason": "diffusion-based photo generation"},
    "huggingface/candle": {"label": 1, "reason": "ML framework in Rust"},
    "LargeWorldModel/LWM": {"label": 1, "reason": "large multimodal world model"},
    "jasonppy/VoiceCraft": {"label": 1, "reason": "neural speech editing/TTS"},
    "apple/corenet": {"label": 1, "reason": "Apple neural-network training library"},
    "Mikubill/sd-webui-controlnet": {"label": 1, "reason": "Stable Diffusion ControlNet extension ('sd' abbreviation defeats 'diffusion' keyword)"},
    "CASIA-LMC-Lab/FastSAM": {"label": 1, "reason": "fast segment-anything vision model"},
    "state-spaces/mamba": {"label": 1, "reason": "sequence-model architecture (LLM building block)"},
    "lllyasviel/Omost": {"label": 1, "reason": "LLM-driven image composition"},
    "facebookresearch/ImageBind": {"label": 1, "reason": "multimodal embedding model"},
    "bigcode-project/starcoder": {"label": 1, "reason": "code LLM"},
    "facebookresearch/seamless_communication": {"label": 1, "reason": "neural speech/text translation models"},
    "pandora-next/deploy": {"label": 1, "reason": "deleted; was a ChatGPT proxy/client (public record)"},
    "zhile-io/pandora": {"label": 1, "reason": "deleted; was a ChatGPT web client (public record)"},
    "systemdesign42/system-design-academy": {"label": 0, "reason": "FALSE POSITIVE: system-design newsletter; incidental 'AI' token in description"}
  },
  "kept_as_is_judgment_calls": {
    "twitter/the-algorithm": {"label": 0, "reason": "recommender system; research.md section 5 explicitly rules it outside the AI-tool rubric (control)"},
    "twitter/the-algorithm-ml": {"label": 0, "reason": "ML models of the same recsys release; kept 0 for consistency with the explicit call above"},
    "adam-maj/tiny-gpu": {"label": 0, "reason": "GPU hardware design education, not AI software"},
    "microsoft/inshellisense": {"label": 0, "reason": "spec-based shell autocomplete, not model-driven"},
    "bleedline/aimoneyhunter": {"label": 1, "reason": "content aggregator ABOUT AI money-making; rubric keyword legitimate"},
    "FujiwaraChoki/MoneyPrinter": {"label": 1, "reason": "automated video tool using AI components; 'chatgpt' keyword legitimate"},
    "harry0703/MoneyPrinterTurbo": {"label": 1, "reason": "LLM-driven video generation; keywords legitimate"},
    "s0md3v/roop": {"label": 1, "reason": "deepfake face-swap; 'diffusion' keyword legitimate"}
  }
}