For most of programming's history, writing code was the expensive act. AI codegen inverted that overnight — and made an old imbalance impossible to ignore.
For decades the expensive act was writing code, so languages optimised the writer: brevity, sugar, keystrokes saved. When agents generate code faster than anyone can read it, that whole economy flips on its head.
What stays scarce — and grows scarcer as volume rises — is knowing the code is correct: the review, the verification, the trust a reader must establish before it ships. Catch a mistake at the source and the saving compounds with every line generated after it.
The gap between the second rung and the third is the entire argument for Vog.
AI floods you with the silent rung specifically — plausible-looking wrong code, faster than review catches it. Vog's entire design converts silent failures into loud ones: unrepresentable where it can, caught at compile time where it must.
Code is read far more than it's written. Rank the audiences by that fact, with trust as the throughline, and every other decision falls out of the order.
Some of the safety comes from active defense; some comes from simply not admitting the feature. Removing expressiveness is a real trade with a real cost — it is made on purpose, and it is worth naming.
The honest caveat: this is a poor trade for a weekend MVP, and a good one for correctness-critical, long-lived, heavily-audited, agent-written systems. Vog knows which it is built for.
Vog is still in design — the repository isn't open yet, but it will be under appropriate open source licenses. No mailing list, no launch date; just the thinking, shared in the open until the code and compiler are ready to share.