AiCompiler – a language where the LLM is the CPU
AiCompiler is a programming language concept where the LLM itself acts as the CPU, allowing developers to write code that is executed directly by a large language model instead of a traditional processor.
Background
- AiCompiler is a programming language/runtime that treats a large language model (LLM) as its execution engine, instead of a traditional CPU. Code written in it is "executed" by an LLM rather than being compiled to machine instructions.
- This is an experimental project, not a mainstream tool — it explores the idea of using LLMs as programmable, general-purpose compute units, blurring the line between code and natural-language prompts.
- It matters because it represents a radical shift in how we think about computation: instead of deterministic, repeatable logic, programs become probabilistic, context-dependent, and more flexible. This has implications for AI-native software, autonomous agents, and self-modifying code.
- Prior context: the rapid rise of LLMs (e.g., GPT, Claude) has led to projects like this that treat LLMs as "soft CPUs" — able to interpret and execute arbitrary instructions in natural language, not just predefined commands.