Contemporary programming languages are optimized for a single reader: the human developer. With the emergence of AI agents as active participants in software construction, verification, and execution, this single-audience assumption has become an architectural liability. We introduce λεξις (Lexis), a language design for dual-audience readability: human and AI with identical semantic fidelity. λεξις fuses four traditions — atomic program construction, geometric glyph density, formal workflow modeling (Petri nets, WWF), and inline proof obligations (dependent types, SMT) — into a unified syntactic frame grounded in a five-level ontological framework. We present ten primitive atoms across three geometric families, a minimal formal grammar, an end-to-end example demonstrating a human-in-the-loop AI agent workflow, and Morphē — a visual canvas that makes λεξις accessible across three levels of user sophistication. We argue that the 1:1 ceremony-to-semantics ratio achieved by λεξις represents a qualitative shift in what programming languages can express, and that the convergence of LLM-assisted proof generation and visual tooling makes this design tractable today.
Tanzi Paolo (Sun,) studied this question.