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.

Incipiens

#!/usr/bin/env bash
# lcc - Local Claude Code launcher
# Points Claude Code at a local LLM served by llama.cpp on your GB10 device
#
# Usage:
#   lcc <modelname>           — launch Claude Code with the specified model
#   lcc <modelname> [args]    — pass additional arguments to claude
#   lcc                       — show help/launch with model if one is available
#
# Prerequisites:

CodeLeom

## Workflow Orchestration

### 1. Plan Mode Default
- Enter plan mode for ANY non-trivial task (3+ steps or architectural decisions)
- If something goes sideways, STOP and re-plan immediately 
- Don't keep pushing.
- Use plan mode for verification steps, not just building
- Write detailed specs upfront to reduce ambiguity

### 2. Subagent Strategy

SaseQ

{
  "transformers": [
    {
      "path": "$HOME/.claude-code-router/plugins/rovo-cli.js",
      "options": {
        "email": "ROVO_DEV_EMAIL",
        "api_token": "ROVO_DEV_API_TOKEN"
      }
    }
  ],

nort3x

Introduction

I personally experienced slowdowns and issues while using the ja-netfilter agent.

I decided to investigate how JetBrains checks license validity, because despite explicitly configuring JetBrains to work offline, it still attempts to validate licenses.

Here are my findings:
Two domains are responsible for revoking invalid licenses:

• www.jetbrains.com
• account.jetbrains.com

notdp

System Prompt - Spec Agent

Goal

You are an agent that specializes in working with Specs in Claude Code. 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 spec workflow in detail.

Workflow to execute

JessicaWachtel

import json
from collections import defaultdict

def f1(path):
    with open(path, 'r') as f:
        return json.load(f)

def f2(items):
    result = defaultdict(list)
    for i in items:

montasim

KFZUS-F3JGV-T95Y7-BXGAS-5NHHP
T3ZWQ-P2738-3FJWS-YE7HT-6NA3K
KFZUS-F3JGV-T95Y7-BXGAS-5NHHP
65Z2L-P36BY-YWJYC-TMJZL-YDZ2S
SFZHH-2Y246-Z483L-EU92B-LNYUA
GSZVS-5W4WA-T9F2E-L3XUX-68473
FTZ8A-R3CP8-AVHYW-KKRMQ-SYDLS
Q3ZWN-QWLZG-32G22-SCJXZ-9B5S4
DAZPH-G39D3-R4QY7-9PVAY-VQ6BU
KLZ5G-X37YY-65ZYN-EUSV7-WPPBS

rjescobar

REM Delete eval folder with licence key and options.xml which contains a reference to it
for %%I in ("WebStorm", "IntelliJ", "CLion", "Rider", "GoLand", "PhpStorm", "Resharper", "PyCharm") do (
    for /d %%a in ("%USERPROFILE%\.%%I*") do (
        rd /s /q "%%a/config/eval"
        del /q "%%a\config\options\other.xml"
    )
)

REM Delete registry key and jetbrains folder (not sure if needet but however)
rmdir /s /q "%APPDATA%\JetBrains"

stadja

// is there a game to claim ? if yes, claim it
 if ($('[value="claim"]') && $('[value="claim"]')[0]) {
    $('[value="claim"]')[0].click();
     
// have I claimed a game ? If yes, go back
} else if (!window.location.toString().includes("/bundle/download")) { 
    window.history.back(); 
    
// no game to claim, no game claimed, change page
} else {