> Blog

Notes on AI-native development and building with LLMs.

by shinpr

Jan 12, 2026 · 9 min

Why LLMs Are Bad at 'First Try' and Great at Verification

Stop optimizing for first-shot perfection. LLMs perform better at verifying and improving existing artifacts than at controlled first-time generation.

#LLM#Agentic Coding#Verification

Jan 12, 2026 · 9 min

Building a Local RAG for Agentic Coding: From Fixed Chunks to Semantic Search with Keyword Boost

Started with a simple RAG for MCP—ended up implementing semantic chunking and rethinking hybrid search. For people who've built RAG systems and started hitting quality limits.

#RAG#Semantic Chunking#MCP

Jan 12, 2026 · 16 min

Rethinking Team Development in the Age of LLMs

Code reviews shift from reviewing code to reviewing specs. Design the whole Spec → Plan → Test → Implementation cycle. Small teams work better—scale by adding teams, not people.

#LLM#Team Development#Process Design