Crazy idea? aiCompiler – write intent in Markdown, LLM executes it as a runtime

- aiCompiler is a new open-source "runtime" that treats Markdown files as executable programs: you write high-level intent in plain text (e.g., "analyze this CSV and generate a chart"), and the system hands it to an LLM (like GPT-4 or Claude) to figure out and run the actual code. - It represents a broader shift toward intent-driven development — instead of writing step-by-step code, the developer specifies the goal, and the AI fills in the implementation details at runtime. - The project lives at aicompiler.dev and has gotten attention on Hacker News and among AI/developer tooling communities, partly because it challenges the assumption that programming languages must be precise and deterministic. - Key context: LLMs are increasingly used not just to *write* code (Copilot, Cursor) but to *execute* it dynamically; aiCompiler is an extreme version of this trend, blurring the line between documentation and software. - Risks and debates: reliability (LLMs can hallucinate the implementation), security (running AI-generated code), and debugging (how do you fix something written as ambiguous prose?).