Hub
AI development: start here
This hub is a curated entry point for practical AI development. The focus is not news or abstractions. It is operational guidance you can apply: how to specify work, how to roll out changes safely, and how to avoid security failure modes.
RAG and LLM operations
- RAG retrieval evaluation checklist
Measure Recall@k, MRR, and nDCG, then separate retrieval failures from generation failures.
- RAG chunking and indexing checklist
Choose chunk boundaries, metadata, and index rules that preserve usable evidence.
- Hybrid search and reranking for RAG
Combine keyword and vector retrieval, then rerank a bounded candidate set.
- LLM evals and regression tests
Build a versioned eval set and block prompt, model, or retrieval regressions.
Foundations
- Spec-to-implementation prompt template
Turn vague requests into implementable specs so AI doesn’t guess the hard parts.
- Clarifying questions checklist
A repeatable way to force constraints, edge cases, and acceptance criteria into the spec.
- Remove AI-tells checklist
Editorial hygiene that makes fast shipping readable and trustworthy.
Security (AI-era failure modes)
- IDOR checklist (authorization)
Object-level authorization rules that survive “vibe-coded” UIs and hidden endpoints.
- IDOR in AI-built web apps
Why AI-built apps regress on authorization, and how to catch it early with tests + patterns.
This hub expands as new RAG, local LLM, LLM ops, and AI security guides ship.