Skip to content

Instantly share code, notes, and snippets.

View jlandure's full-sized avatar
💭
👋

Julien Landuré jlandure

💭
👋
View GitHub Profile
@jlandure
jlandure / gist:3175119
Last active October 7, 2015 13:57
maven command line
mvn -Dmaven.test.skip
mvn dependency:resolve
mvn dependency:tree : arbre
mvn install:install-file -Dfile=as3corelib.swc -DgroupId=com.adobe -DartifactId=as3corelib -Dversion=0.93 -Dpackaging=swc -DgeneratePom=true -DcreateChecksum=true
#compiler src/test/java car compile tout court ne fait que src/main/java
mvn test-compile
mvn versions:set -DnewVersion=1.2.3-SNAPSHOT
@jlandure
jlandure / gist:3368187
Created August 16, 2012 07:50
Shortcuts
#notepad++
indent XML : CRLT + ALT + SHIFT + B
#eclipse
##code
CTRL + espace : completion
CTRL + 1 : correcton auto
CTRL + SHIFT + O : organize import
CTRL + SHIFT + F : formatter le code
CTRL + I : indenter
@hgwood
hgwood / walk.js
Last active September 6, 2016 14:14
Walking a directory tree synchronously with Node.js 6+
const fs = require('fs');
function walk(root) {
const files = [];
const dirs = [root];
while (dirs.length) {
const dir = dirs.shift();
const dircontent = fs.readdirSync(dir).map(file => path.join(dir, file));
const [newDirs, newFiles] = partition(dircontent, file => fs.statSync(file).isDirectory());
files.push(...newFiles);
@hsablonniere
hsablonniere / README.adoc
Last active May 11, 2024 12:50
Conference bingo

Conference bingo

Background

In many tech conferences, attendees are invited to rate the talk and/or the speaker from 1 to 5 stars. This type of ratings is interesting but has a few drawbacks.

The discussion started as a twitter thread with this french proposition.

C9M xEZWsAARmav

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.