Rafael Santos Sá fael

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.

A series of posts on css-in-js

0. styles as objects

First, an exercise. Can we represent all of css with plain data? Let's try.

let redText = { color: 'red' };

Code is clean if it can be understood easily – by everyone on the team. Clean code can be read and enhanced by a developer other than its original author. With understandability comes readability, changeability, extensibility and maintainability.

General rules

Follow standard conventions.
Keep it simple stupid. Simpler is always better. Reduce complexity as much as possible.
Boy scout rule. Leave the campground cleaner than you found it.
Always find root cause. Always look for the root cause of a problem.

Design rules

From zero to microservice with 𝚫 now

The following guide will show you how to deploy a simple microservice written in JavaScript using 𝚫 now.

It uses Open Source tools that are widely available, tested and understood:

Node.JS
NPM
Express

	//@version=4
	study(title="CTO Line", shorttitle="CTO", overlay=true, resolution="")
	smma(src, length) =>
	smma = 0.0
	smma := na(smma[1]) ? sma(src, length) : (smma[1] * (length - 1) + src) / length
	smma
	v1 = smma(hl2, 15)
	m1 = smma(hl2, 19)
	m2 = smma(hl2, 25)
	v2 = smma(hl2, 29)

	import * as path from "https://deno.land/std@0.75.0/path/mod.ts";
	import * as fs from "https://deno.land/std@0.75.0/fs/mod.ts";
	import { paramCase } from "https://deno.land/x/case@v2.1.0/mod.ts";

	const cwd = Deno.cwd();

	// Files to be excluded during the walk
	const skip = (await Deno.readTextFile(path.join(cwd, ".gitignore")))
	.replace(/#.*/g, "")
	.replace(/\s*\n/g, "\n")

	import { useState, useMemo, useEffect } from "react";
	import { interpret } from "xstate/lib/interpreter";

	export default function useMachine(machine, { log = false } = {}) {
	const [current, setCurrent] = useState(machine.initialState);

	const service = useMemo(() => interpret(machine), []);

	useEffect(() => {
	service.onTransition(setCurrent).start();

	export class MyComponent extends React.Component {
	componentWillMount() {
	const styleid = 'MyComponentStyles'
	if (document.head.querySelector(`#${styleid}`) !== null) {
	const styles = document.createElement('style');
	styles.id = styleid;
	styles.append('[foo]{ color: blue; }');
	document.head.appendChild(styles);
	}
	}

	import React, { Component } from 'react';
	import PropTypes from 'prop-types';
	import { connect } from 'react-redux';

	import {
	extractClockTimestamp,
	} from './reduxClockReducer';

	import {
	ACTION_CLOCK_SUBSCRIBE,

	/*
	New Relic GPT ad monitoring code
	//
	This snippet captures PageActions containing references to Google publisher tag ads
	Should you have any questions about this, please reach out to me at rmusser@newrelic.com
	Please note that you must be using New Relic's SPA browser agent.
	//
	*/
	//
	//first, make sure the newrelic object exists properly