ShikChen

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.

Tool / Project	Type / Architecture	UI / Dashboard	CLI / API	Multi‑node / Clustering	Auto TLS / SSL / ACME	Zero‑downtime / Rolling Deploy	Preview / PR Environments	Team / Permissions	Volume / Backup Support	Open Source / License	Heroku‑style Compatibility	Pros / Strengths	Cons / Caveats / Weaknesses
Dokku	Single‑server (extensions for multi)	Minimal (no central UI, though plugins exist)	CLI, plugin APIs	Via community / plugin / custom setups	Yes (Let’s Encrypt plugin)	Via Docker techniques / plugin	Limited, requires extra setup	Basic (single admin; community plugins)	Plugin / manual (e.g. volume backups)	MIT (fully open source)	Good: git push, buildpack style, plugins add DB, etc.	Very stable, mature, lots of community extensions	Scaling / multi‑node is manual; UI is weak; managing many apps / teams is harder
Dokploy	Multi‑node / cluster (Docker Swarm)	Yes (built‑in UI)	CLI + AP

hi! this is a list of all¹ the paths in the splatnet 3 application as of september 10th, 2022. a lot of these are probably not useful for external applications. i found all these manually through developer tools. some may be incorrect. a : indicates that it's a parameter. i have marked what i considered to be notable with a ⭐️ emoji.

these can be used to launch the splatnet 3 application on a device with the nintendo switch online app installed. on ios, this is done with com.nintendo.znca://znca/game/4834290508791808?p=<url encoded path>. i am not sure how this works on android, but it's probably similar. for example, to open the anarchy schedule, use com.nintendo.znca://znca/game/4834290508791808?p=%2Fschedule%2Fbankara.

SUS Format Specification v2.7 (rev2)

注: SUS は今や Sliding Universal Score の略であり、 SeaUrchin Score ではありません。

1. 概要

テキストデータであり、全て printable character で構成される。
拡張子: *.sus
EOL: CRLF もしくは LF
エンコード: 常に UTF-8
# で始まる行がデータとして意味を持ち、それ以外の行はコメントとして無視される。
文字列データを指定する部分は " ～ " でクォートする。

CUSTOM GAME ENGINES: A Small Study

WARNING: Article moved to separate repo to allow users contributions: https://github.com/raysan5/custom_game_engines

A couple of weeks ago I played (and finished) A Plague Tale, a game by Asobo Studio. I was really captivated by the game, not only by the beautiful graphics but also by the story and the locations in the game. I decided to investigate a bit about the game tech and I was surprised to see it was developed with a custom engine by a relatively small studio. I know there are some companies using custom engines but it's very difficult to find a detailed market study with that kind of information curated and updated. So this article.

Nowadays lots of companies choose engines like [Unreal](https:

	"""
	The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
	This file is the complete algorithm.
	Everything else is just efficiency.

	@karpathy
	"""

	import os # os.path.exists
	import math # math.log, math.exp

	"""
	The 2024 Transformer (the Noam Transformer):
	- RMSNorm
	- GQA or some combination
	- Sliding window attention
	- Swiglu
	- RoPE (Rotary Positional Embedding)

	LLM Arches:
	hidden \| MLP mult. \| n_layers \| rope_theta \| GQA Group Size \| GLU Act. \| ops

	/*
	================================================================================
	modified from this: https://github.com/its-a-feature/macos_execute_from_memory (supports only bundle)
	code injection : https://github.com/CylanceVulnResearch/osx_runbin by Stephanie Archibald (does not support m1 x64 emulation and FAT header)

	added FAT header (universal Macho) parsing

	script-kiddied, debugged, etc. by @exploitpreacher
	================================================================================
	*/

	#include <stdint.h>
	#include <stdio.h>
	#include <stdlib.h>

	// munged from https://github.com/simontime/Resead

	namespace sead
	{
	class Random
	{

	template <int MAXV, class T = int> struct Dinic {
	const static bool SCALING = false; // non-scaling = V^2E, Scaling=VElog(U) with higher constant
	int lim = 1;
	const T INF = numeric_limits<T>::max();
	struct edge {
	int to, rev;
	T cap, flow;
	};
	int s = MAXV - 2, t = MAXV - 1;