Vibe coding with AI

A behind-the-scenes look at Auranite.com, from the original idea to the PHP backend, moderation workflow, media pipeline, and interactive world map.

Why OpenAI's GPT-5.6 Sol matters less as a faster autocomplete model and more as an agent that can carry engineering work to done.

A practical model-routing guide for choosing OpenAI's Sol, Terra, or Luna by task risk, complexity, speed, and cost.

What Anthropic's Claude Fable 5 says about long-horizon coding, asynchronous work, visual checks, and the changing role of the developer.

A grounded way to compare GPT-5.6 Sol and Claude Fable 5 using your own tasks, acceptance criteria, review effort, and cost per accepted change.

AI coding has moved from suggesting the next line to operating editors, terminals, browsers, and cloud environments. Here's what that changes.

AI app builders make polished prototypes quickly, but recent research shows why backend integrity, security, and concurrency still need deliberate engineering.

Why files, repository rules, examples, tools, and feedback loops now matter more than finding one magical prompt.

How to use parallel coding agents, worktrees, clear ownership, and an integration pass without multiplying conflicts and confusion.

Token prices are only the beginning. Measure retries, review, cleanup, latency, and accepted changes to understand the real economics of AI coding.