veekaybee

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

jboner

Latency Comparison Numbers (~2012)
----------------------------------
L1 cache reference                           0.5 ns
Branch mispredict                            5   ns
L2 cache reference                           7   ns                      14x L1 cache
Mutex lock/unlock                           25   ns
Main memory reference                      100   ns                      20x L2 cache, 200x L1 cache
Compress 1K bytes with Zippy             3,000   ns        3 us
Send 1K bytes over 1 Gbps network       10,000   ns       10 us
Read 4K randomly from SSD*             150,000   ns      150 us          ~1GB/sec SSD

rondy

FWIW: I (@rondy) am not the creator of the content shared here, which is an excerpt from Edmond Lau's book. I simply copied and pasted it from another location and saved it as a personal note, before it gained popularity on news.ycombinator.com. Unfortunately, I cannot recall the exact origin of the original source, nor was I able to find the author's name, so I am can't provide the appropriate credits.

---

Effective Engineer - Notes

• By Edmond Lau
• Highly Recommended 👍
• http://www.theeffectiveengineer.com/

What's an Effective Engineer?

posener

Story: Writing Scripts with Go

This is a story about how I tried to use Go for scripting. In this story, I’ll discuss the need for a Go script, how we would expect it to behave and the possible implementations; During the discussion I’ll deep dive to scripts, shells, and shebangs. Finally, we’ll discuss solutions that will make Go scripts work.

Why Go is good for scripting?

While python and bash are popular scripting languages, C, C++ and Java are not used for scripts at all, and some languages are somewhere in between.

bishboria

These links no longer work. Springer have pulled the free plug.

Graduate texts in mathematics

duplicates = multiple editions

A Classical Introduction to Modern Number Theory, Kenneth Ireland Michael Rosen

A Classical Introduction to Modern Number Theory, Kenneth Ireland Michael Rosen

debasishg

General Background and Overview

1. Probabilistic Data Structures for Web Analytics and Data Mining : A great overview of the space of probabilistic data structures and how they are used in approximation algorithm implementation.
2. Models and Issues in Data Stream Systems
3. Philippe Flajolet’s contribution to streaming algorithms : A presentation by Jérémie Lumbroso that visits some of the hostorical perspectives and how it all began with Flajolet
4. Approximate Frequency Counts over Data Streams by Gurmeet Singh Manku & Rajeev Motwani : One of the early papers on the subject.
5. [Methods for Finding Frequent Items in Data Streams](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.187.9800&rep=rep1&t

trevnorris

The prep-script.sh will setup the latest Node and install the latest perf version on your Linux box.

When you want to generate the flame graph, run the following (folder locations taken from install script):

sudo sysctl kernel.kptr_restrict=0
# May also have to do the following:
# (additional reading http://unix.stackexchange.com/questions/14227/do-i-need-root-admin-permissions-to-run-userspace-perf-tool-perf-events-ar )
sudo sysctl kernel.perf_event_paranoid=0

grownseed

Give me back my sanity

One of the many things I do for my group at work is to take care of automating as many things as possible. It usually brings me a lot of satisfaction, mostly because I get a kick out of making people's lives easier.

But sometimes, maybe too often, I end up in drawn-out struggles with machines and programs. And sometimes, these struggles bring me to the edge of despair, so much so that I regularly consider living on a computer-less island growing vegetables for a living.

This is the story of how I had to install Pandoc in a CentOS 6 Docker container. But more generally, this is the story of how I think computing is inherently broken, how programmers (myself included) tend to think that their way is the way, how we're ultimately replicating what most of us think is wrong with society, building upon layers and layers of (best-case scenario) obscure and/or weak foundations.

*I would like to extend my gratitude to Google, StackOverflow, GitHub issues but mostly, the people who make the

mrdoob

General

• https://aframe.io/aframe-inspector/example/
• https://aws.amazon.com/sumerian/
• https://hologram.cool/
• https://janusvr.com/
• https://nunustudio.org/
• http://ottifox.com/
• https://patches.vizor.io/
• https://playcanvas.com/

mattdesl

optimize GLTF file

This optimizes a GLTF file that was exported by blender (or similar) by de-duplicating buffer views (i.e. chunks of bytes) that are equal and removing redundant accessors.

For example, 100 cubes of different scales/materials/rotations/etc should all end up using a single BufferGeometry in ThreeJS, which isn't the case with current GLTF exporters in Blender and parsers for ThreeJS.

In scenes with a lot of instancing, it can dramatically reduce total file size as well as render performance. In one test scene:

Before: 4.8MB file size, 832 THREE.Geometry instances across 832 THREE.Mesh objects
After: 661KB file size, 13 THREE.Geometry instances across 832 THREE.Mesh objects