Llm9p: LLM as a Plan 9 file system

Meet llm9p, the throwback-meets-future hack that lets you talk to an AI by writing to a file and reading the reply—no fancy apps, just your computer’s everyday file tricks. It uses the old-school 9P network file system (think: treat faraway stuff like local files) to hook into Large Language Models (LLMs), with support for Anthropic’s API, the Claude Code command-line tool, and local models coming soon. The vibe? Retro wizardry colliding with modern AI hype—and everyone has thoughts.

Commenter mleroy frames it as part of a bigger movement: “files as the ideal interface for agents,” linking to a lively HN thread. Fans gush that this makes AI “scriptable like Lego,” chaining prompts with pipes and shell commands. But then the plot twist: stingraycharles drops a reality check, asking how llm9p sets a “system prompt” with Claude Code when the tool doesn’t truly support it—apparently “—appendSystemPrompt” just posts a normal user message. Cue the “is this brilliant or cursed?” debate. Jokes fly about “cat-ing a robot” and “pipelines building Rube Goldberg AIs,” while skeptics mutter “just use curl.” It’s nostalgia vs practicality, hacker joy vs confusing edge cases. Whether genius or gimmick, the crowd agrees: it’s delightfully chaotic, and very on-brand for the Plan 9 fan club and shell-script maximalists.