Andrey Shmelev maestrow

TokenZip — PRD, HLD, LLD

📋 PRD — Product Requirements Document

1. Executive Summary

TokenZip v2 transforms Karpathy's llm wiki concept into a gzip like token compression engine on top of entire codebase, which can reduce the LLM input token cost upto by 95% when using with Coding Copilots like Claude Code, Codex etc. Instead of generating a flat text summary, it builds a multi-level, queryable, chainable knowledge graph — from repo → modules → files → symbols — stored locally in .tokenzip/db, exposed as an MCP server for any AI copilot, and kept fresh via git hooks

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.

Nix vs Guix

These are notes to the stream: https://youtu.be/S9V-pcTrdL8

Some notes

We are not aware of a lot of GNU software available to us.
Seems that Guix more hacker-friendly/explorable.

General comparsion

Description

Nix

Guix

Comment

Ubuntu/Debian Offline Repository Creation Gist

When I googled how to create my own offline repository of packages for use in an offline Ubuntu/Debian machine, the results were disheartening and the steps to be taken scattered all over the place.

The files within this gist will allow you to:

Download specific apt-get packages... with dependencies included!
Create a Packages.gz file so that you can add the repository folder you create to the target machine's /etc/apt/sources.list file.

	https://medium.com/@clem.boin/creating-a-minimal-kernel-development-setup-using-qemu-and-archlinux-987896954d84

	# Install Arch system
	qemu-image -f qcow2 kernel-dev-archlinux.img 4G
	wget http://mirrors.edge.kernel.org/archlinux/iso/2018.12.01/archlinux-2018.12.01-x86_64.iso
	# Note that ping does not work here
	qemu-system-x86_64 -cdrom archlinux-2018.12.01-x86_64.iso -boot order=d -drive file=kernel-dev-archlinux.img,format=qcow2 -m 2G -enable-kvm -cpu host -smp 8 -net user,hostfwd=tcp::10022-:22 -net nic

	pct create <id> /var/lib/vz/template/cache/centos-7-default_20170504_amd64.tar.xz \
	-arch amd64 \
	-ostype <centos\|ubuntu\|etc> \
	-hostname <hostname> \
	-cores <cores> \
	-memory <memory(MB)> \
	-swap <swap(MB)> \
	-storage local-lvm \
	-password \
	-net0 name=eth0,bridge=<bridge>,gw=<gateway>,ip=<cidr>,type=veth &&\

	package main

	import (
	"crypto/rand"
	"encoding/base64"
	"fmt"
	"io"
	"math/big"
	)

	open FSharp.Quotations

	let rec getMethodInfo = function
	\| Patterns.Call(_,``method``,_) -> ``method``
	\| Patterns.Lambda(_,body) -> getMethodInfo body
	\| _ -> failwith "Unexpected Form"

	let getGenericMethodInfo functionExpression =
	let methodInfo = getMethodInfo functionExpression
	if methodInfo.IsGenericMethod then

	namespace Example1

	open WebSharper
	open WebSharper.UI.Next
	open WebSharper.UI.Next.Html
	open WebSharper.UI.Next.Client

	[<JavaScript>]
	module Counter =

	open System
	open System.Windows.Forms
	open System.Drawing

	type Action =
	\| Increment
	\| Decrement


	let form = new Form(Width= 400, Height = 300, Visible = true, Text = "Hello World")