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.

mberman84

git clone https://github.com/OpenDevin/OpenDevin.git
cd OpenDevin
conda create -n od python=3.10
conda activate od
docker ps
(optional) install docker if not already installed
docker pull ghcr.io/opendevin/sandbox
export OPENAI_API_KEY={your key}
(optional I had to install rust) curl --proto '=https' --tlsv1.2 -sSf [https://sh.rustup.rs](https://sh.rustup.rs/) | sh
(optional) restart terminal

jossef

import requests
from requests.adapters import HTTPAdapter, Retry


def get_vscode_extensions(max_page=10000, page_size=100,
                          include_versions=True, include_files=True, include_category_and_tags=True, include_shared_accounts=True, include_version_properties=True,
                          exclude_non_validated=False, include_installation_targets=True, include_asset_uri=True, include_statistics=True,
                          include_latest_version_only=False, unpublished=False, include_name_conflict_info=True, api_version='7.2-preview.1', session=None):
    if not session:
        session = requests.session()

veekaybee

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

candideu

Hello world!

As a video editor, researcher, digital media enthusiast, and lover of all things FLOSS, I've been on the hunt for an open source alternative to proprietary services like Otter.ai, Sonix, and Descript. I've pitched my idead on open-source-ideas, but I wanted to create a dedicated post for it so that it can reach as many people as possible.

Project description

The idea

A simple, easy-to-use application where users can dictate or upload audio or video files, and an automated transcript is generated. This transcript is synced to the audio track, clickable, and editable, so that users can skip to certain passages and refine the transcript accordingly.

The revised transcript can then be exported as plain text, .srt caption file (and other subtitle formats), .pdf, shareable web page, etc. for further processing.

kaaquist

Podman with docker-compose on MacOS.

Podman an alternative to Docker Desktop on MacOS

Getting podman installed and started is super easy.
Just use brew to install it.

> brew install podman

Now since podman uses a VM just like the Docker Client on MacOS we need to initialize that and start it.

rapatil

ozaki-r

#!/bin/sh

if [ $# -lt 2 ]; then
        echo "usage: $0 name zone [time]"
        exit 1
fi

# TODO: check if authenticated and suggest 'gcloud auth login'

name=$1

qrtt1

ITHome 第 12 屆鐵人賽

499978920

# path
# C:\Program Files\PowerShell\7

function set_proxy {
	$proxy = 'http://127.0.0.1:7890'

    # temporary
    $env:HTTP_PROXY = $proxy
    $env:HTTPS_PROXY = $proxy