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.

pmarreck

MFIC — Mechanically-Falsifiable Independent Control

A discipline for checks that can't be fooled by the same mistake that produced the work.

Abstract. When the author of a piece of work — an LLM, or a human moving fast — can be confidently and silently wrong, the usual defenses fail quietly: a test written from the same mistaken assumption as the code passes while the code is broken. MFIC is the class of check that escapes this trap, defined by four properties that must all hold at once. It is Mechanical (cases swept by machine, not hand-picked, so nothing is silently omitted), Falsifiable (each case genuinely bites when the work is wrong, on inputs you couldn't pre-arrange to pass), Independent (its verdict comes from the contract or the data — a source the producer doesn't control — which is segregation of duties restated for software), and a real Control (it holds the authority to block the bad outcome, not merely log it). This is the COSO / Sarbanes-Oxley model of internal co

Shrutiind02

AI Humanizer	Pangram Labs	Quetext	Best For	Verdict
WriteHuman	100% Human ✅	96% Human (Low AI likelihood)	AI detector bypass	The only tool that successfully bypassed both Pangram Labs and Quetext, making it the best performer for bypassing AI detectors.
GPTHuman	100% AI	80% Human (Low AI likelihood)	Natural writing	Produced the most natural-sounding output while preserving the original meaning and formatting. Successfully bypassed Quetext but not Pangram Labs.
Undetectable AI	100% AI	72% AI (High AI likelihood)	Flexible writing modes	May bypass some AI detectors, but not consistently. Premium Stealth Mode may perform better than the free versio

thomasnield

import pandas as pd
import torch

"""
Complete the code below to perform stochastic gradient descent
on a linear regression.

Try to find a sufficient learning rate and number of iterations. 
"""

slightfoot

// MIT License
//
// Copyright (c) 2026 Simon Lightfoot
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:

urholaukkarinen

Related to: ValveSoftware/steam-for-linux#10722

On Linux (X11), when you receive a Steam chat message, the chat window does not open. The window is actually created behind the scenes, but it is never shown.

• fixsteamchat.sh script listens to Steam logs, detects when a chat window is created, and forces it to show.
• fixsteamchat.service can be used to turn the script into a systemd service that runs in the background.

The script requires xdotool: sudo apt install xdotool

artemgetmann

🧠 How to Save Context Tokens When Using Claude Code

This is a personal reference workflow for minimizing token usage while maintaining project continuity across Claude Code (Sonnet 4 with file access).

---

✅ Setup: Populate CLAUDE.md

Claude loads CLAUDE.md automatically at session start.

thomasnield

import pandas as pd
import torch

"""
COMPLETE THE CODE BELOW BY REPLACING THE QUESTION MARKS ?'s
SO FORWARD PROPAGATION IS COMPLETE 
"""


url = "https://raw.githubusercontent.com/thomasnield/machine-learning-demo-data/master/classification/maintenance_predict.csv"

ocornut

Memory Optimization (Christer Ericson, GDC 2003)
http://realtimecollisiondetection.net/pubs/GDC03_Ericson_Memory_Optimization.ppt

Cache coherency primer (Fabian Giesen)
https://fgiesen.wordpress.com/2014/07/07/cache-coherency/

Code Clinic 2015: How to Write Code the Compiler Can Actually Optimize (Mike Acton)
http://gdcvault.com/play/1021866/Code-Clinic-2015-How-to