Sergey Egorov egorse

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.

Summary of book "Clean Architecture" by Robert C. Martin

Uncle Bob, the well known author of Clean Code, is coming back to us with a new book called Clean Architecture which wants to take a larger view on how to create software.

Even if Clean Code is one of the major book around OOP and code design (mainly by presenting the SOLID principles), I was not totally impressed by the book.

Clean Architecture leaves me with the same feeling, even if it's pushing the development world to do better, has some good stories and present robust principles to build software.

The book is build around 34 chapters organised in chapters.

Story: Writing Scripts with Go

This is a story about how I tried to use Go for scripting. In this story, I’ll discuss the need for a Go script, how we would expect it to behave and the possible implementations; During the discussion I’ll deep dive to scripts, shells, and shebangs. Finally, we’ll discuss solutions that will make Go scripts work.

Why Go is good for scripting?

While python and bash are popular scripting languages, C, C++ and Java are not used for scripts at all, and some languages are somewhere in between.

security and hardening options for systemd service units

A common and reliable pattern in service unit files is thus:

NoNewPrivileges=yes
PrivateTmp=yes
PrivateDevices=yes
DevicePolicy=closed
ProtectSystem=strict

QEMU arm64 cloud server emulation

This is basically a rehash of an original post on CNXSoft - all credit (particularly for the Virtio device arguments used below) belongs to the author of that piece.

Download the latest uefi1.img image. E.g. ubuntu-16.04-server-cloudimg-arm64-uefi1.img from https://cloud-images.ubuntu.com/releases/16.04/release/

Download the UEFI firmware image QEMU_EFI.fd from https://releases.linaro.org/components/kernel/uefi-linaro/latest/release/qemu64/

Determine your current username and get your current ssh public key:

	"""
	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

	<!DOCTYPE html>
	<html lang="en">

	<head>
	<meta charset="UTF-8">
	<title>Vulnerability Report</title>

	<!-- Template Metadata-->
	<meta name="author" content="OnceUponALoop">
	<meta name="version" content="1.0.0">

	#!/bin/bash

	# REPO_NAME=<repo>.git
	# ORIGIN_URL=git@<host>:<project>/$REPO_NAME
	# REPO1_URL=git@<host>:<project>/$REPO_NAME

	rm -rf $REPO_NAME
	git clone --bare $ORIGIN_URL
	if [ "$?" != "0" ]; then
	echo "ERROR: failed clone of $ORIGIN_URL"

	@virgo:~ $ sudo tcpdump -vvvv -ttt -i eth1 icmp6 and 'ip6[40] = 134'
	00:00:00.000000 IP6 (hlim 255, next-header ICMPv6 (58) payload length: 200) fe80::b675:eff:fefa:1cb > ip6-allnodes: [icmp6 sum ok] ICMP6, router advertisement, length 200
	hop limit 0, Flags [managed, other stateful], pref medium, router lifetime 1800s, reachable time 0s, retrans time 0s
	source link-address option (1), length 8 (1): b4:75:0e:fa:01:cb
	0x0000: b475 0efa 01cb
	mtu option (5), length 8 (1): 1280
	0x0000: 0000 0000 0500
	prefix info option (3), length 32 (4): 2a01:ffff:43f::/64, Flags [onlink, auto], valid time 7200s, pref. time 1800s
	0x0000: 40c0 0000 1c20 0000 0708 0000 0000 2a01
	0x0010: ffff 043f 0000 0000 0000 0000 0000

	# Basic Strongswan ikev2 server setup

	* paltform: atlantic.net ubuntu 14.04 x64
	* the commands below are run with root account

	## Strongswan
	```
	apt-get install strongswan
	apt-get install iptables iptables-persistent
	```