VZVlad Zoffincodapult.hashnode.dev·Oct 1 · 11 min readThe Agent Shouldn't Be Able to Approve Its Own RulesWhy policy changes need a different boundary from code changes when AI agents work on real projects. One of the less obvious problems I've run into with coding agents isn't that they make bad changes.04FC
VZVlad Zoffincodapult.hashnode.dev·Sep 25 · 11 min readA Dependency Graph Shows Structure. It Doesn't Show Intent.Why understanding a codebase requires more than knowing which modules depend on which. A dependency graph is one of the first things I wanted when I started thinking about tooling for AI coding agents00
VZVlad Zoffincodapult.hashnode.dev·Sep 23 · 10 min readThe Tests Passed. The Architecture Didn't.One of the stranger things about working with coding agents is that a change can be completely valid and still be wrong. TypeScript passes. The tests pass. The build is green. The feature works. And I19RBM
VZVlad Zoffincodapult.hashnode.dev·Aug 12 · 5 min readThe Next.js SaaS Boilerplate Checklist for 2026Most SaaS boilerplates look similar on the homepage: auth, payments, dashboard, emails, blog, and a fast launch promise. The differences appear later, when a customer asks for team roles, billing chan00
VZVlad Zoffincodapult.hashnode.dev·Aug 10 · 7 min readShipFast vs Supastarter vs MakerKit vs Codapult: Choosing a SaaS Starter in 2026SaaS starters are easy to compare badly. The cheapest way is to count pages, components, or screenshots. That misses the part that decides whether the starter still helps after the first launch: how m00
