AI Engineer's Workflow Analysis | Generated by AI
Here’s the full analysis:
================================================================
REPO & USAGE AUDIT
================================================================
STATUS: All 117 repos from repos.json are ALREADY cloned locally.
Your ~/projects/ has 305 repos total — 117 tracked, 186 misc/personal.
================================================================
YOUR TOP TOOLS (shell history)
================================================================
546 ww ← your daily driver
85 hermes
46 git
46 uv
12 tmux
6 zed
6 ssh
3 codex
Top ww subcommands:
414 ww note ← note-taking is your #1 workflow
45 ww git
20 ww projects
16 ww amd-dev-cloud
9 ww hf
9 ww github
================================================================
YOUR DOCK APPS
================================================================
WeChat, Safari, Telegram, Discord, Microsoft To Do,
Ghostty, Alacritty, Warp, Zed, GitHub Desktop,
System Settings, Passwords, Clock
================================================================
OBSERVATIONS
================================================================
1. You're deep in AI engineering (agents, training, inference, models)
2. Note-taking is your #1 daily activity (414 ww note calls)
3. You use 3 terminals: Ghostty, Alacritty, Warp
4. You do GPU work (amd-dev-cloud, ROCm)
5. You care about AI agents (hermes, codex, claw-code, opencode)
Based on your usage patterns, here are repos you’re missing that would fit your workflow:
================================================================
RECOMMENDED NEW REPOS TO CLONE
================================================================
--- AI Agents & Coding (your biggest interest) ---
anthropics/claude-code # Claude Code CLI — you use codex already
anthropics/anthropic-cookbook # Official Anthropic recipes/notebooks
openai/openai-cookbook # OpenAI recipes — practical LLM patterns
langchain-ai/langgraph # Agent graph framework — industry standard
crewAIInc/crewAI # Multi-agent orchestration
Significant-Gravitas/AutoGPT # Pioneering autonomous agent
mem0ai/mem0 # Long-term memory for AI agents
--- LLM Training & Research (you train on H200/RTX 4070) ---
huggingface/trl # SFT/DPO/GRPO — you have fine-tuning skill
huggingface/peft # LoRA/QLoRA — already in peft-fine-tuning skill
vllm-project/vllm # Already cloned but HIGH PRIORITY for you
NVIDIA/Megatron-LM # Already cloned ✓
pytorch/torchtitan # PyTorch native distributed training
pytorch/ao # PyTorch architecture optimization (quantization)
state-spaces/mamba # Already cloned ✓
--- LLM Inference & Serving ---
ggml-org/llama.cpp # Already cloned ✓
oobabooga/text-generation-webui # Popular local LLM webui
Mozilla-Ocho/llamafile # Single-file LLM binaries — very portable
NexaAI/nexa-sdk # Local model serving
--- Developer Tools & CLI (you live in terminal) ---
jesseduffield/lazygit # TUI for git — you use git 46x
sharkdp/bat # Better cat with syntax highlighting
sharkdp/fd # Better find
BurntSushi/ripgrep # Better grep (you already use via hermes)
ajeetdsouza/zoxide # Smarter cd (you cd 107x)
astral-sh/ruff # Python linter (already in your pyproject.toml)
sxyazi/yazi # Terminal file manager
charmbracelet/glow # Markdown renderer in terminal
dandavison/delta # Better git diff
--- AI Research Papers & Notebooks ---
karpathy/minGPT # Minimal GPT implementation (Karpathy)
karpathy/ng-video-lecture # Companion code for Karpathy's lectures
microsoft/autogen # Already cloned ✓
deepseek-ai/DeepSeek-V3 # Already cloned ✓
--- Vector DBs & RAG (you have read/ RAG module) ---
chroma-core/chroma # Already cloned ✓
qdrant/qdrant # Already cloned ✓
weaviate/weaviate # Popular vector DB — missing
run-llama/llama_index # RAG framework — missing
langchain-ai/langchain # LLM framework — missing (despite langchain-test dir)
--- Monitoring & Observability ---
langfuse/langfuse # Already cloned ✓
BerriAI/litellm # Already cloned ✓
Top 10 I’d clone first (highest signal for your workflow):
1. langchain-ai/langgraph — agent graphs, industry standard
2. jesseduffield/lazygit — you use git constantly
3. ajeetdsouza/zoxide — smarter cd (you cd 107x)
4. anthropics/claude-code — you already use codex
5. huggingface/trl — GRPO/DPO fine-tuning
6. run-llama/llama_index — RAG framework
7. pytorch/torchtitan — distributed training
8. Mozilla-Ocho/llamafile — portable LLM binaries
9. karpathy/minGPT — Karpathy-style minimal impl
10. crewAIInc/crewAI — multi-agent framework
Want me to clone any of these? Just say which ones (or "clone all top 10").