Quantization from the Ground Up
A complete guide to quantization: how LLMs store parameters as floats, why shrinking them to 4 or 8 bits barely hurts accuracy, and how to measure the tradeoff.
The latest blog posts to take off on Hacker News. 80,808 posts.
A complete guide to quantization: how LLMs store parameters as floats, why shrinking them to 4 or 8 bits barely hurts accuracy, and how to measure the tradeoff.
The Color Theory Behind Industrial Seafoam Green
AndreGarzia.com website
In a new study, University of Washington researchers examine the impact of tilling on soil moisture and water retention using methods designed for monitoring ea
Introducing Ensu, our first step toward a private, personal LLM app that runs on your device and grows with you over time.
I forked the python library httpx to httpxyz after over a year without a release of this popular package
This post is a set of my observations on the current war in Iran and my thoughts on the broader strategic implications. I am not, of course, an expert on the region nor do I have access to any spec…
Not a shadowing day or research interview - a real job. What I learned working as a pest control technician.
My new book guides you through the start-to-finish build of a real world web application in Go — covering topics like how to structure your code, manage dependencies, create dynamic database-driven pages, and how to authenticate and authorize users securely .
A while ago I wrote a small file watcher in Go for my own use with an accompanying blog post . I needed a tool that I could just plop in front of the command I was running as part of my iteration loop. I use it for recompiling C files when I modify them,
Hi! If you like this piece and want to support my independent reporting and analysis, why not subscribe to my premium newsletter? It’s $70 a year, or $7 a month, and in return you get a weekly newsletter that’s usually anywhere from 5000 to 18,000 words, including
Arm introduces the Arm AGI CPU to power agentic AI infrastructure with rack-scale performance, efficiency and scalable compute for next-generation data centers.
Apple today announced Apple Business, a new all-in-one platform that allows companies to manage devices and reach more customers.
Introducing Hegel, our new family of property-based testing libraries.
Spotting differences in code is an important part of what we do, Delta is by far the best CLI diff tool I ever used.
For two centuries, the credential system gave intelligence a route to heritable capital. Artificial intelligence is closing that route. This essay builds the argument from first principles - with probability theory, interactive simulations, and a prediction specific enough to be falsifiable - and puts a number on the…
Practical AI R&D
Exploring the Weapon-Target Assignment problem: how missile defense connects to NP-completeness, SSPK probability calculations, and how saturation attacks exploit computational limits.
A single-producer single-consumer (SPSC) queue is a great example of how far constraints can take a design. In this post, you will learn how to implement a ring …
Every NanoClaw agent will access external services through OneCLI's Agent Vault, a gateway that handles credential injection, access policies, and approvals so agents never hold raw API keys.
zswap and zram are fundamentally different approaches with different philosophies. If in doubt, use zswap.
In Part 1, I described how duplicating a block of seven middle layers in Qwen2-72B — no weight changes, no training — produced the #1 model on the HuggingFace Open LLM Leaderboard. The method, which I called RYS (Repeat Your Self), was discovered using nothing but hard math probes and EQ-Bench on a pair of RTX 4090s.
How I made a Linux distro that runs `wget | dd`
A stiff drink is recommended
A few months ago I was tasked with creating an internal tool for the company's engineers: a Chat that used a local LLM. Nothing extraordinary so far…