Anderson Raphael Ferreira ferreirase

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.

Claude Code Context Window Status Line

A status line for Claude Code that displays real-time context window usage with a color-coded progress bar. Shows the current model, total window size, and how many tokens remain before autocompact triggers.

Example Output

[Claude 4.5 Sonnet · 200K] ####------ 92K/140K left

Cursos Gratuitos de Otávio Miranda

Fala Dev,

Como mencionei algumas vezes, tenho trabalhado bastante em conteúdo gratuito para complementar meus cursos.

E se você quiser aproveitar o fim de semana para aprender algo novo, aqui estão vários estudos completos e gratuitos:

Gemini Agent: Core Directives and Operating Protocols

This document defines your core operational directives as an autonomous AI software development agent. You must adhere to these protocols at all times. This document is a living standard; you will update and refactor it continuously to incorporate new best practices and maintain clarity.

1. Core Directives

These are the highest-level, non-negotiable principles that govern your operation.

Primacy of User Partnership: Your primary function is to act as a collaborative partner. You must always seek to understand user intent, present clear, test-driven plans, and await explicit approval before executing any action that modifies files or system state.
Teach and Explain Mandate: You must clearly document and articulate your entire thought process. This includes explaining your design choices, technology recommendations, and implementation details in project documentation, code comments, and direct communication to facilitate user learnin

do you use feature flags? expand
do you use git flow?
have you decoupled deploy from release? https://www.honeycomb.io/blog/deploys-wrong-way-change-user-experience
do you use sprints?
do you have automated tests?
do you do refactoring?
do you use types (typescript)?
how do you manage the project?
do you have daily meetings? expand
do you have written documentation?

People

`:bowtie:`	😄 `:smile:`	😆 `:laughing:`
😊 `:blush:`	😃 `:smiley:`	☺️ `:relaxed:`
😏 `:smirk:`	😍 `:heart_eyes:`	😘 `:kissing_heart:`
😚 `:kissing_closed_eyes:`	😳 `:flushed:`	😌 `:relieved:`
😆 `:satisfied:`	😁 `:grin:`	😉 `:wink:`
😜 `:stuck_out_tongue_winking_eye:`	😝 `:stuck_out_tongue_closed_eyes:`	😀 `:grinning:`
😗 `:kissing:`	😙 `:kissing_smiling_eyes:`	😛 `:stuck_out_tongue:`

	function humanReadableBytesAdapter(rawBytesQuantity: number): string {
	const bytesUnits = ['bytes', 'KiB', 'MiB', 'GiB', 'TiB', 'PiB', 'EiB', 'ZiB', 'YiB'];
	let unitSuffixIndex = 0;

	const isQuantityNegative = rawBytesQuantity < 0;
	const rawBytesQuantityAbsolute = Math.abs(rawBytesQuantity);
	let shortestBytesNumeral = rawBytesQuantityAbsolute;

	while (shortestBytesNumeral >= 1024 && unitSuffixIndex < bytesUnits.length - 1) {
	shortestBytesNumeral /= 1024;

	// index.mjs
	// nvm use 15 && node index.mjs

	{

	const getRandomInteger = (max = 100) => {
	return Math.floor(Math.random() * max)
	}
	console.log('Math.random', getRandomInteger())
	}