Field note 02LinkedSpec / 2026

Why AI Needs a New Programming Language

A language for machine authors does not need to explain every step to a person. It needs to preserve intent while removing avoidable work.

AI can write JavaScript, Rust, Python, and every other language we give it. That is useful, but it is not the same as designing a language for an intelligence that can reason over structure directly.

Translation is an expensive habit

When an AI receives an outcome and returns human-oriented source code, it is performing two jobs: deciding what should happen and packaging that decision in a format designed for a human maintainer. The second job consumes tokens, context, network traffic, and time.

A new language can preserve the first job while changing the second. It can express the constraints that matter, leave implementation choices open, and produce a traceable execution plan for review.

Designed around intent

LinkedSpec is an experiment in that direction. It is not a claim that humans disappear from software. It is a proposal that human judgment belongs at the level of outcomes and boundaries, while machines handle the translation to the most efficient runtime form.

← Back to field notes