// blog
Notes from the AI trenches.
System thinking, rescue stories, and the things we keep getting asked — in Kunjan’s voice. Anti-hype, senior, and practical.
2026-06-20 · 6 min read
Where vibe-coded code breaks (and how to spot it before it ships)
AI-generated code fails in predictable ways. Here are the five failure modes I see in almost every rescue, and how to catch them before they become an outage.
AI Architecture Code Review
Read →
2026-06-12 · 7 min read
RAG without the bankruptcy: architecting LLM systems that stay affordable
Most RAG systems are architected for the demo, then surprise everyone with the bill. Here's how to design retrieval, evals, and cost in from the start.
AI RAG Architecture
Read →
2026-06-04 · 5 min read
Why I rewrote nothing: the case against the big rewrite
When a codebase is a mess, the rewrite is tempting and almost always wrong. Here's why I refactor incrementally instead, and how a rescue actually works.
Architecture Refactoring Engineering
Read →