Sunny Gonnabathula sunny-g

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.

Beads Uninstall Script

A comprehensive uninstall/cleanup script for Beads (bd) that removes all traces of the tool from a system.

Usage

./scripts/uninstall.sh            # dry-run (scan $HOME)
./scripts/uninstall.sh --apply    # perform cleanup
./scripts/uninstall.sh --root DIR --apply

GitHub Pages File-Sharing

If you sometimes find yourself needing to share a file over HTTP, there are not many file-sharing solutions you can use.

This action works by publishing files via GitHub Pages behind an obscure prefix path. It updates the repository README with the URLs, so that you have links to the published files.

You can keep the repository secret, and the files are public but semi-secret, so you can share them without concern that other files will be discovered.

UNRAID NFS Configuration Guide - Complete Issue Resolution

Executive Summary

This document details the investigation, root cause analysis, and resolution of recurring NFS stale file handle issues in a home lab environment with UNRAID NAS and multiple Linux clients. The solution involved migrating from static NFS mounts to systemd automount configuration, eliminating stale handle problems while maintaining full container compatibility.

Environment Overview

Infrastructure

UNRAID Server: unraid-server (192.168.1.100) - Primary NAS with NFS exports

Making MLX Go Fast

This guide assumes you have some familiarity with MLX and want to make your MLX model or algorithm as efficient as possible.

The guide covers the following:

Graph Evaluation
Type Promotion
Operations

The Interaction Calculus

The Interaction Calculus (IC) is term rewriting system inspired by the Lambda Calculus (λC), but with some major differences:

Vars are affine: they can only occur up to one time.
Vars are global: they can occur anywhere in the program.
There is a new core primitive: the superposition.

An IC term is defined by the following grammar:

	#!/usr/bin/env node

	// SupVM.ts
	// ========
	// A tiny evaluator for the HVM subset used by InsertGen_hardcoded.hvm.
	// Sup-fork captures are the only DUP form needed by that file. SupVM gives
	// them the same observable branch correlation through a labelled-choice map,
	// and charges captured variables as DUP-like interactions.

	declare function require(nam: string): any;

	"""
	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

	<div style="position:absolute;top:-999px;left:-999px">
	<svg
	id="effectSvg"
	width="200"
	height="200"
	viewBox="0 0 200 200"
	xmlns="http://www.w3.org/2000/svg">


	<filter id="displacementFilter4">

	import Control.Monad (when)
	import Data.Char (chr, ord)
	import Data.IORef
	import Data.Word
	import Debug.Trace
	import System.IO.Unsafe (unsafePerformIO)
	import Text.Parsec hiding (State)
	import qualified Data.IntMap.Strict as IntMap
	import qualified Data.Map as Map
	import qualified Text.Parsec as Parsec