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.

daniel-farina

Running OpenCode with Gemma 4 26B on macOS (via llama.cpp)

As of April 2026, Gemma 4 tool calling is broken in Ollama v0.20.0 (ollama/ollama#15241) - the tool call parser fails and streaming drops tool calls entirely. OpenCode also has issues with local OpenAI-compatible providers (anomalyco/opencode#20669, #20719).

This guide documents a working setup using:

• llama.cpp (built from source with PR #21326 template fix + PR #21343 tokenizer fix) instead of Ollama
• OpenCode built from source with PR #16531 tool-call compatibility layer

Tested on macOS Apple Silicon (M1 Max, 32GB) on April 2, 2026.

alesha-pro

I want you to help me make requests (prompts) for the Stable Diffusion neural network.

Stable diffusion is a text-based image generation model that can create diverse and high-quality images based on your requests. In order to get the best results from Stable diffusion, you need to follow some guidelines when composing prompts.

Here are some tips for writing prompts for Stable diffusion1:

1) Be as specific as possible in your requests. Stable diffusion handles concrete prompts better than abstract or ambiguous ones. For example, instead of “portrait of a woman” it is better to write “portrait of a woman with brown eyes and red hair in Renaissance style”.
2) Specify specific art styles or materials. If you want to get an image in a certain style or with a certain texture, then specify this in your request. For example, instead of “landscape” it is better to write “watercolor landscape with mountains and lake".
3) Specify specific artists for reference. If you want to get an image similar to the work of some

bluelovers

Stable Diffusion is an AI art generation model similar to DALLE-2.
Here are some prompts for generating art with Stable Diffusion. 

Example:

- A ghostly apparition drifting through a haunted mansion's grand ballroom, illuminated by flickering candlelight. Eerie, ethereal, moody lighting. 
- portait of a homer simpson archer shooting arrow at forest monster, front game card, drark, marvel comics, dark, smooth
- pirate, deep focus, fantasy, matte, sharp focus
- red dead redemption 2, cinematic view, epic sky, detailed, low angle, high detail, warm lighting, volumetric, godrays, vivid, beautiful
- a fantasy style portrait painting of rachel lane / alison brie hybrid in the style of francois boucher oil painting, rpg portrait

cb109

/**
 * The yup library has no builtin way to check if items in an array
 * match one or more defined shapes, we can only check against a set of
 * whitelisted values using yup.array().oneOf([..]).
 *
 * Using yup.addMethod() and yup.mixed().test() however we can pretty
 * much add any custom validation we want. The function below allows to
 * make validation pass for array items of different shapes.
 */

2357gi

--- Actions ---
$Copy                                              <M-C>
$Cut                                               <M-X> <S-Del>
$Delete                                            <Del> <BS> <M-BS>
$LRU                                              
$Paste                                             <M-V>
$Redo                                              <M-S-Z> <A-S-BS>
$SearchWeb                                        
$SelectAll                                         <M-A>
$Undo                                              <M-Z>

bonedaddy

add
block/get
block/put
block/stat
cat
cid/base32
cid/bases
cid/codecs
cid/format
cid/hashes

fireattack

#!/usr/bin/env python
# vim:fileencoding=UTF-8:ts=4:sw=4:sta:et:sts=4:ai

import sys

import os, getopt, struct
import imghdr

def get_image_type(imgname, imgdata=None):
    imgtype = imghdr.what(imgname, imgdata)

oliveratgithub

{
	"emojis": [
		{"emoji": "👩‍👩‍👧‍👧", "name": "family: woman, woman, girl, girl", "shortname": ":woman_woman_girl_girl:", "unicode": "1F469 200D 1F469 200D 1F467 200D 1F467", "html": "👩‍👩‍👧‍👧", "category": "People & Body (family)", "order": ""},
		{"emoji": "👩‍👩‍👧‍👦", "name": "family: woman, woman, girl, boy", "shortname": ":woman_woman_girl_boy:", "unicode": "1F469 200D 1F469 200D 1F467 200D 1F466", "html": "👩‍👩‍👧‍👦", "category": "People & Body (family)", "order": ""},
		{"emoji": "👩‍👩‍👦‍👦", "name": "family: woman, woman, boy, boy", "shortname": ":woman_woman_boy_boy:", "unicode": "1F469 200D 1F469 200D 1F466 200D 1F466", "html": "👩‍👩‍👦‍👦", "category": "People & Body (family)", "order": ""},
		{"emoji": "👨‍👩‍👧‍👧", "name": "family: man, woman, girl, girl", "shortname": ":man_woman_girl_girl:", "unicode": "1F468 200D 1F469 200D 1F467 200D 1F467", "html": "👨‍👩&z

Tithen-Firion

// Usage: phantomjs openload.js <video_url>
// if that doesn't work try: phantomjs --ssl-protocol=any openload.js <video_url>

var separator = ' | ';
var page = require('webpage').create(),
  system = require('system'),
  id, match;

if(system.args.length < 2) {
  console.error('No URL provided');