Blog
Shipping fast with AI, without breaking what works.
Case studies, real problems and practical playbooks for developers, CTOs and agencies working with Claude Code, Cursor, Codex and Copilot.
Case study: an AI agent broke checkout during a one-line copy change
A button-text change shipped a broken checkout. Here's how, and the three guardrails that stop it.
6 min read →ProblemWhy your AI coding agent keeps bringing back bugs you already fixed
Agents forget why odd code exists. A decision log and locks give them a memory.
5 min read →PlaybookTaking over a codebase written by AI: a CTO's first week
A five-day playbook for inheriting a vibe-coded app without breaking it.
7 min read →ProblemRunning several AI coding agents on one repo without collisions
Parallel agents and teammates on one repo: scopes, branches and checks that keep them apart.
6 min read →Case studyCase study: a dev agency shipping client work with AI agents, safely
How a small agency stopped AI regressions on client apps, and sells the guardrails.
6 min read →TemplateA post-mortem template for regressions caused by AI coding agents
A blameless template for AI-caused bugs, and how to turn each one into a guardrail.
5 min read →Founder noteWhy I built Stet
AI agents made me faster, then broke things that already worked.
4 min read →Building in publicWhat our first outside tester taught us
What worked, the three things they asked for, and how all three shipped.
4 min read →