Will Benton willb

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.

How to get firmware onto mutable modules

I had the best luck with the Black Magic Probe and gdb.

Install compiler

First install the arm compiler:

sudo add-apt-repository ppa:team-gcc-arm-embedded/ppa

## temporary installing Strimzi
oc apply -f https://github.com/strimzi/strimzi-kafka-operator/releases/download/0.8.2/strimzi-cluster-operator-0.8.2.yaml
# Simple Cluster with one ZK and one Kafka node
oc apply -f https://raw.githubusercontent.com/strimzi/strimzi-kafka-operator/master/examples/kafka/kafka-persistent-single.yaml

oc get pods -n myproject -w

wait...

Ok, I geeked out, and this is probably more information than you need. But it completely answers the question. Sorry. ☺

Locally, I'm at this commit:

$ git show
commit d6cd1e2bd19e03a81132a23b2025920577f84e37
Author: jnthn <jnthn@jnthn.net>
Date:   Sun Apr 15 16:35:03 2012 +0200

When I added FIRST/NEXT/LAST, it was idiomatic but not quite so fast. This makes it faster. Another little bit of masak++'s program.

	import altair as alt
	from vega_datasets import data

	source = alt.topo_feature(data.world_110m.url, "countries")
	osm_url, otm_url = ('https://tile.openstreetmap.org/', 'https://tile.opentopomap.org/')
	select_urls = alt.binding_select(options=[osm_url, otm_url], name='select tile service')
	param_urls = alt.param(bind=select_urls, value=osm_url)
	param_tx = alt.param(expr="width / 2")
	param_ty = alt.param(expr="height / 2")
	param_base_tile_size = alt.param(value=256)

	#!/bin/sh

	while getopts ":if:" opt; do
	case $opt in
	f)
	useFile=$OPTARG
	;;
	\?)
	echo "Invalid option: -$OPTARG" >&2
	exit 1

	import argparse

	from time import clock

	from pyspark.sql import SparkSession
	from pyspark.mllib.clustering import KMeans
	from numpy import fromstring

	parser = argparse.ArgumentParser()
	parser.add_argument('--master', help='Spark master URL (default: "local[]")', default="local[]")


	// 'small' was a subsample of meta-full parquet
	scala> small.orderBy(small("@timestamp")).map(utils.toJSON).saveAsTextFile("/home/eje/logdemo/logs_ordered.json")