AI-assisted development

A practical build order grounded in one real, disclosed case: one engineer, one week, then thousands of weekly users. What to do, and in what order.

I built a real feature with an AI agent and counted the exact tokens. The real cost ran 6.3x past what counting the code alone would suggest.

A December 2025 study of 470 real GitHub pull requests found AI-co-authored code carries 1.7x more issues, concentrated in logic and security findings.

A real database deleted mid-freeze, hallucinated packages with 30,000 downloads, a prompt-injected agent: four real failures, and the fix for each.

Karpathy coined vibe coding in 2025, then coined agentic engineering a year later. What changed, and which term actually describes what you're doing.

Auto Mode is now Claude Code's default for Pro, Max, and Team plans. Here's what it actually blocks, how to configure it, and the real security tradeoff.

GitHub Copilot and Cursor both work fine for two people on individual plans. Here is the real seat cost math, and when the team tier is worth paying for.

Product engineer isn't a new title, but AI coding agents are reshaping what it requires day to day. Agentic engineer is the newer role forming alongside it.

Prompt, context, loop, and harness engineering optimize four different layers of an AI agent. Here's what separates them, and what a real harness is made of.

AI coding agents are breaking sprint planning and story points. The one Scrum artifact that gets harder to skip is the one everyone already ignored.

Most 'best MCP servers' lists are unverified directories. Here's a short, hands-on-tested shortlist for everyday dev work, and what didn't make the cut.

AI isn't replacing senior engineers, it's changing what the job is: the real shift from writing code to directing and supervising the agents that write it.

MCP servers aren't just another API layer. Here's what actually happens when a coding agent calls one, with a real setup, not the spec restated.

Vibe coding isn't inherently bad, unreviewed output shipped somewhere it matters is. Here's exactly what breaks when nobody checks it, and why.

We migrated the same 60-file API through Claude Code, Cursor, and GitHub Copilot. Here's what each one missed, and which one we'd trust with production code.

What to commit, what to gitignore, and why the CLAUDE.md, Cursor rules, or AGENTS.md file drifts out of date faster than anything else in the repo.