paperthin
The skillbook for recursive self-improvement (RSI). NVIDIA's latest proof, unpacked with timeless engineering. Distilled from the OpenAI harness engineering hac
What is paperthin?
The skillbook for recursive self-improvement (RSI). NVIDIA's latest proof, unpacked with timeless engineering. Distilled from the OpenAI harness engineering hackathon winner's zero-touch method.
Overview of paperthin
What it is
Paperthin is a skillbook for recursive self-improvement (RSI). It provides low-level agentic design patterns distilled from NVIDIA's latest proof and the OpenAI harness engineering hackathon winner's zero-touch method. It is designed to work on any agent, including Codex, Claude Code, OpenCode, Cursor, Grok Build, Grok Bot, Muse, and Pi.
Key features
- Offers over 30 skills organized into depth, breadth, and coil categories for various agentic tasks.
- Includes skills for rewriting, debloating, ordering, detooling, style fixing, and more.
- Provides read-only skills for checking understanding, suggesting next steps, and questioning decisions.
- Supports installation via plugin marketplace commands or a one-line npx command for most agents.
- Features a zero-touch method that updates skills automatically at session start for some agents.
Use cases
Paperthin can be used for recursive self-improvement workflows, such as rewriting messy documents, trimming bloated code, checking facts against real sources, and planning work cycles. It is suitable for developers and teams using AI agents to improve codebases, documentation, and decision-making processes.
Getting started
See the repository README for detailed installation instructions. Quickstart commands for specific agents are provided there.
License and pricing
The license is MIT. The pricing type is open source.
Based on the project's public README and repository metadata.