Show HN: A prompt that builds the most capable AI agent system

Copy‑paste a “super AI” that runs everything—commenters ask: genius or just a mega wall of text

TLDR: A Show HN post shares a colossal prompt claiming to build a self‑improving AI agent that can do everything from coding to company ops. Commenters push back: it’s too long and unclear, debating one giant prompt versus modular skills—important because people want practical, trustworthy tools, not hype.

A bold Show HN drop claims you can paste one giant prompt into your favorite coding agent and—boom—get a self‑improving “agentic operating system” that can code, research, automate browsers and desktops, and even help run a company. It promises plans, skills, evaluations, and learning loops across tools like Claude Code, Cursor, and OpenHands. Cue the comment section lighting up like a holiday tree. The first vibe check: disbelief at the sheer size. As one early reaction put it: that prompt is enormous and the ambition even bigger.

Then the split: modularists vs monolith lovers. One frustrated voice wonders why a copy‑paste novel isn’t delivered as a tidy set of skills you can slot in, instead of a single mega spell. Skeptics hint at “vaporware vibes,” noting the doc tells the agent to build a system but doesn’t ship actual code. Optimists counter: if it really scaffolds, verifies, and learns, maybe giant instructions are the shortcut we’ve been waiting for. Meanwhile, jokes fly fast—“TL;DR for the TL;DR,” “War and Peace as a prompt,” and the instant classic: copy‑paste your CEO. Whether this is the ultimate productivity hack or just the longest love letter to agents ever written, the community is united on one thing: it’s a lot to swallow, and everyone wants receipts.

Key Points

  • A single, long prompt is provided to make a coding agent build a self-improving agentic operating system for computer-based work.
  • The system targets tasks across software engineering, research, operations, analysis, customer support, sales/marketing ops, and multi-month projects.
  • The prompt can wrap an existing agent with a harness (plans, skills, evals, self-improvement loops) or build a new system via an agent SDK.
  • Quick start: copy-paste the prompt into supported tools (e.g., Claude Code, OpenAI Codex, Cursor, OpenHands, Claude Agent SDK) to begin building immediately.
  • A strict protocol emphasizes measurable results, transparent state, closed-loop execution, minimal questions, artifact creation, and continuous adherence checks.

Hottest takes

“That’s a very big prompt” — jonwinstanley
“Wouldn’t it make more sense to build it as a set of skills?” — ValentineC
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.