Vog is a language design project built around the cost of being wrong. Correct code is cheap in the long run: you pay for it once. Loud failure is bounded: a compile error, red test, or rejected build tells you where to look while the cause is still nearby. Silent failure is the expensive one. It ships looking fine, reaches production, and sends the bill later, detached from the line that caused it.
AI coding agents make that ladder matter more. The problem is not only that they can produce plausible code quickly; plausible wrong code is already making it to production at an alarming rate. Vog’s design response is to move as many mistakes as possible down the ladder: from silent to loud, from loud to unrepresentable. That means static types, immutability by default, visible Ok/Ng failure, no null, no truthiness, no exceptions, no metaprogramming layer where behavior can hide, tests that carry their red or green state in the source, and tooling for the conventions that should not become compiler rules.
The goal is not magic, and it is not brevity. Vog spends keystrokes to buy confidence: readers before writers, clarity before convenience, and ordinary mistakes made louder, rarer, and easier to find.