Jay Zheng
notes on building with AI
---
LinkedIn
GitHub
jay.j.zheng-at-gmail

now
---
• exploring MCP servers and agent tooling
• building semantic code search with vector embeddings
• reading about scaling laws and in-context learning
• building a Claude Code-like agent from scratch to understand the agentic loop

projects
---
• VectCode github
  semantic code search via vector embeddings + MCP server
  Go / ChromaDB / Ollama / MCP

• HAC github
  HTTP Agent Context — safety metadata spec for AI agents interacting with APIs
  RFC-style spec / JSON Schema

• Football Data
  NFL data pipeline + ML projections + LLM panel deliberation system
  Go / Python / SQLite / scikit-learn / Claude API / React

reading
---
• Attention Is All You Need
• Scaling Laws for Neural Language Models
• RAG for Knowledge-Intensive NLP Tasks
• ReAct: Synergizing Reasoning and Acting
• Toolformer: Language Models Can Teach Themselves to Use Tools
• Constitutional AI: Harmlessness from AI Feedback
• Model Context Protocol spec
• Context Engineering 2.0
---
• The Art of Unix Programming formatted pdf
• Representational State Transfer
• Communicating Sequential Processes