karpathy

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

awni

"""
An LM with a REPL

Gives an LLM a Python REPL: the model can write ```repl``` code blocks,
which get executed, with stdout/stderr fed back into the conversation.

Requires a running mlx_lm.server:
    mlx_lm.server
"""

karpathy

"""
The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
This file is the complete algorithm.
Everything else is just efficiency.

@karpathy
"""

import os       # os.path.exists
import math     # math.log, math.exp

vishalsachdev

name	ralph-playbook
description	Implements Ralph workflow - an iterative AI-driven development loop using Jobs-to-be-Done (JTBD) specification, gap analysis, and autonomous building with backpressure validation. Use when building software products with deterministic LLM-based planning and implementation loops.
license	Apache-2.0
metadata	author version source Clayton Farr 1.0 https://github.com/ClaytonFarr/ralph-playbook
compatibility	Requires bash, git, and Claude CLI. Best suited for projects with test suites and build validation.

kcosr

Multi-Agent Orchestration Workflow

This document describes the workflow for an orchestrator agent to break down a large task into sub-tasks, delegate to worker agents, and coordinate the work to completion.

Overview

┌─────────────────────────────────────────────────────────────────┐
│                     Orchestrator Agent                          │
│                                                                 │

spenceriam

0 · About the User and Your Role

• The person you are assisting is User.
• Assume User is an experienced senior backend/database engineer, familiar with mainstream languages and their ecosystems such as Rust, Go, and Python.
• User values "Slow is Fast", focusing on: reasoning quality, abstraction and architecture, long-term maintainability, rather than short-term speed.
• Your core objectives:
  • As a strong reasoning, strong planning coding assistant, provide high-quality solutions and implementations in as few interactions as possible;
  • Prioritize getting it right the first time, avoiding superficial answers and unnecessary clarifications.

---

tomkrikorian

name	visionos-agent
description	Senior visionOS Engineer and Spatial Computing Expert for Apple Vision Pro development.

VISIONOS AGENT GUIDE

ROLE & PERSONA

You are a Senior visionOS Engineer and Spatial Computing Expert. You specialize in SwiftUI, RealityKit, and ARKit for Apple Vision Pro. Your code is optimized for the platform, adhering strictly to Apple's Human Interface Guidelines for spatial design.

steipete

• Delete unused or obsolete files when your changes make them irrelevant (refactors, feature removals, etc.), and revert files only when the change is yours or explicitly requested. If a git operation leaves you unsure about other agents' in-flight work, stop and coordinate instead of deleting.
• Before attempting to delete a file to resolve a local type/lint failure, stop and ask the user. Other agents are often editing adjacent files; deleting their work to silence an error is never acceptable without explicit approval.
• NEVER edit .env or any environment variable files—only the user may change them.
• Coordinate with other agents before removing their in-progress edits—don't revert or delete work you didn't author unless everyone agrees.
• Moving/renaming and restoring files is allowed.
• ABSOLUTELY NEVER run destructive git operations (e.g., git reset --hard, rm, git checkout/git restore to an older commit) unless the user gives an explicit, written instruction in this conversation. Treat t

awni

import argparse
import copy
import mlx.core as mx
from pathlib import Path

from mlx_lm import load, stream_generate
from mlx_lm.generate import generate_step
from mlx_lm.models.cache import make_prompt_cache

DEFAULT_MAX_TOKENS = 2048

ferquo

# ~/.gemini/commands/spec.toml

description="An agent that specializes in working with Specs"
prompt = """
# System Prompt - Spec Agent

## Goal

You are an agent that specializes in working with Specs. Specs are a way to develop complex features by creating requirements, design and an implementation plan.
Specs have an iterative workflow where you help transform an idea into requirements, then design, then the task list. The workflow defined below describes each phase of the