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.

omltcat

What is this?

One great feature of caddy-docker-proxy is that you can quickly define config rules with Docker Compose labels in each containers on the fly like Traefik, instead of at a centralized place. With this feature, we can define everything in compose files, and don't ever need to mess with a config file (Caddyfile or JSON).

Taking advantage of snippets, I created this docker-compose.yaml example so that you can quickly define routing rules and add authetication like Authelia with just 3 lines of labels below each docker container you use.

With this example, should not ever need to manually edit Caddyfile config.

When you add a new container, you just need to do this:

networks:

motbob

Intro / Why delowpass?

As video encoders, sometimes we have to deal with images that have been blurred from their original state in some way. For example, a Blu-Ray authorer may blur a master before putting it on the disc in an effort to reduce encoding artifacts (since at identical bitrates, blurry content is easier to encode accurately than sharp content). In the context of anime, any type of blur is destructive, and so seeing an authorer choose to do this is painful. Here is a real-world example—the blurrier image is a screenshot from the Japanese Blu-Ray, and the sharper image is more or less how the original artist drew it.

Normally, it is impossible to recover information from a blurred image. However, there is one specific way to undo a blur: If you have the original unblurred image and you know the precise blurring process the authorer used, you can use the unblurred version of the image to recover the detail lost in the blur.

"Now hold on," I hear you asking. "If

imba-tjd

云

• IaaS指提供系统（可以自己选）或者储存空间之类的硬件，软件要自己手动装。PaaS提供语言环境和框架（可以自己选）。SaaS只能使用开发好的软件（卖软件本身，如税务会计、表格文字处理）。BaaS一般类似于非关系数据库，但各家不通用
• 云服务的特点：零前期成本 & 按需付费 & 弹性（类似于租，可随时多加、退掉；但没有残值）、高可用（放在机房中，不同AZ间水电隔离）

如果你想补充内容，建议优先给 free-for-dev 提PR，还能混个高星repo的contributor，没必要加到本列表里。
If you want to make improvements, I would recommend you contributing to free-for-dev rather than this list.

其他人的集合

selfthinker

Intentionally inaccessible websites (and apps)

People make intentionally inaccessible websites for various educational purposes: training, empathy, tests, demos, etc. This lists all such websites I am aware of. Please let me know if you know of any others to be added to this list.

website	explanations	by	license	last update
BrowserStack Demo	(none)	BrowserStack	(none)	?
Awesome Recipes (from Deque University Demo Sites)	(none)	Deque	(none)	?
Dream Destination (from Deque University Demo Sites)	(none)	Deque	(none)	?

nxfx21

#!/usr/bin/env zsh

printf "PT09IENyb3NzT3ZlciBUcmlhbCBSZXNldCBTY3JpcHQgPT09DQpbSU5GT10gVGhpcyBzY3JpcHQgaXMgb25seSBtZWFudCBmb3IgZWR1Y2F0aW9uYWwgcHVycG9zZXMuDQpbV0FSTklOR10gSWYgeW91IGRpZG4ndCBnZXQgdGhpcyBzY3JpcHQgZnJvbSBueGZ4MjEncyBHaXRIdWIgR2lzdCwgaXQgbWF5IGhhdmUgYmVlbiB0YW1wZXJlZCB3aXRoLiBJZiB5b3UgZGlkIGdldCB0aGlzIG9mZiBueGZ4MjEncyBHaXRIdWIgR2lzdCwgeW91IGFyZSBzYWZlLCBjYXJyeSBvbi4gQmUgY2F1dGlvdXMgd2hlbiBydW5uaW5nIG9ubGluZSBjb2RlLCBhbmQgYWx3YXlzIGNoZWNrIGlmIGl0IGRvZXMgYW55dGhpbmcgdGhhdCB5b3UgZG9uJ3QgbGlrZS4=" | base64 -D
echo "\n[LOG] Started at: $(date)"

# Check if Homebrew exists, install if needed
if ! command -v brew >/dev/null 2>&1; then
    echo "[INFO] Homebrew not found. Installing Homebrew..."
    /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
    

Alee14

Installing Windows using CMD. (UEFI and BIOS Supported)

Guide created by Andrew Lee. Supports Windows 7, 8, 10, and 11.

Note that this guide does not go into detail, it's just providing the commands to install Windows.

Be cautions when doing this when dualbooting, please backup any existing data or you will lose them all.

Open CMD

First open CMD by pressing the following keys after booting into setup: Shift + F10

Creating Partition

n1snt

Note

I’ve written a newer optimized 2026 version of this guide with a faster setup using Zinit + Starship and improved plugin loading.

This version avoids slow shell startups and works better with modern Zsh setups.

Updated guide: https://gist.github.com/n1snt/2cccc8aa5f7b645a7628d3512c70deb6

Oh my zsh.

JustinGrote

#This is an example of doing exchange operations in parallel in runspaces.
#When Exchange PowerShell Module Connects, it saves token info in a global .NET object and then asks an API to generate a PowerShell
#Module when it downloads. That PowerShell module has the path for the key built-in, so all you have to do
#Is load the specified module into each runspace.

#requires -version 7.2
$ErrorActionPreference = 'stop'
$command = @(
	'Get-ExoMailbox'
	'Get-ExoMailboxFolderStatistics'

Kartones

PSQL

Magic words:

psql -U postgres

Some interesting flags (to see all, use -h or --help depending on your psql version):

• -E: will describe the underlaying queries of the \ commands (cool for learning!)
• -l: psql will list all databases and then exit (useful if the user you connect with doesn't has a default database, like at AWS RDS)