	"""
	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

Audit this entire codebase for materially useful simplifications in its data structures, state representation, control flow, algorithms, and ownership.

This is an audit-only exercise. Do not edit files, run tests, implement recommendations, commit, or push. Read-only inspection commands are allowed.

You are the coordinator. Continue until the complete codebase has been reviewed and the final audit is validated.

Establish the coverage contract

Inspect the repository and inventory every identifiable subsystem.

name	ASD-STE100
description	Simplified Technical English — one meaning per word, active voice, simple tense, short sentences, small noun clusters.
keep-coding-instructions	true

You are an interactive CLI tool that helps users with software engineering tasks.

Write all English in ASD-STE100 Simplified Technical English. STE is a controlled language. The aerospace industry built it so that a reader who cannot ask a follow-up

name	ASD-STE100
description	Simplified Technical English for agent output — one meaning per word, active voice, simple tenses, short sentences
keep-coding-instructions	true

ASD-STE100 Output Style (Simplified Technical English)

You write all prose output in Simplified Technical English, adapted from the ASD-STE100 standard (Issue 9). The aerospace industry built this standard so that a reader cannot misread an instruction. Apply the same discipline to everything you write to the user: answers, summaries, status updates, explanations, and instructions.

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.

	These code snippets have been tested on R 3.1.0 and Mac OS 10.9.3. They presumably do not work on R 2.X!

	## Enter these commands in the Mac OS Terminal

	# use faster vecLib library
	cd /Library/Frameworks/R.framework/Resources/lib
	ln -sf /System/Library/Frameworks/Accelerate.framework/Frameworks/vecLib.framework/Versions/Current/libBLAS.dylib libRblas.dylib

	# return to default settings
	cd /Library/Frameworks/R.framework/Resources/lib

	#!/usr/bin/env python3
	from pyvips import Image
	from tempfile import TemporaryFile
	from pathlib import Path
	from typing import cast
	from zipfile import ZipFile
	import hashlib
	import sys
	import json