Blank turns your TV screen black until you press any button on a remote
I’m happy to announce the release of a new tvOS app called Blank. It turns your screen black and keeps it that way until you press any button on a remote. Seriously, that’s all it does. Here’s the screen you see when you launch the app for the first time: That second paragraph hints at […]
Microsoft forced people to install their browser (2010)
It was a happy time for PC users. Microsoft had just delivered Windows 95, an incredibly vast improvement over windows 3.1 and DOS.
What to do when you’ve ruined your life
Every now and then, life becomes unacceptable. What do you do then?
Large language models: Scaling laws and emergent properties
More details on the science of training large autoregressive language models.
High Court rules that Getty vs. Stability AI case can proceed
While most of the attention regarding artificial intelligence and copyright has been centred in the United States, it’s useful to remember that one of the main pieces of litigation is taking …
Steve Jobs and the actually usable computer (2011)
At a sad time at which the world has just lost Steve Jobs, it is well to reflect on some of the things which he, his passion, and his creations have given us, as now the technical community should take...
About That OpenAI "Breakthrough"
Will OpenAI’s Q* change the world?
Development slowness in big and legacy applications
Software companies come in many sizes. There are small startups, medium-sized companies, and huge enterprises. As you might expect, startups are usually lean and fast. Big companies that develop large applications move much more slowly. Those might be systems developed for many years or decades by hundreds of…
Which kinds of GPT startups will thrive?
How to separate the wheat from the chaff when GPT-powered startups are popping up everywhere
Imagination GPUs now support OpenGL 4.6
Since its launch in 1992 OpenGL® has been used extensively by software developers for PCs and workstations to create high-performance, visually compelling graphics applications.
How unions won a 30% raise for every fast food worker in California
Anything that can't go on forever will eventually stop. 40 years of declining worker power shattered the American Dream (TM), producing multiple generations whose children fared worse than their parents, cratering faith in institutions and hope for a better future.
Transcoding Unicode with AVX-512: AMD Zen 4 vs. Intel Ice Lake
Most systems today rely on Unicode strings. However, we have two popular Unicode formats: UTF-8 and UTF-16. We often need to convert from one format to the other. For example, you might have a database formatted with UTF-16, but you need to produce JSON documents using UTF-8. This conversion is often called…
Deteriorating mitochondria linked to major depression in older adults
'One problem feeds into another, and make what began as a small issue into a much larger one'
Cut out everything that's not surprising (2019)
Derek Sivers official site. Thoughts on philosophy, culture, self-improvement. Author of Useful Not True, How to Live, Hell Yeah or No, Anything You Want.
Earth's blobs are remnants of an ancient planetary collision
In the 1980s, geophysicists made a startling discovery: Two continent-sized blobs of unusual material were found deep near the center of the Earth, one beneath the African continent and one beneath the Pacific Ocean.Each blob is twice the size of the moon, and research over the last decade has shown that they are…
The Linux kernel has been accidentally hardcoded to a maximum of 8 cores
TL;DR This doesn’t mean that the scheduler can’t use more than 8 cores. The scheduler controls how to allocate tasks to available cores. How to schedule particular workloads efficiently…
Pharmacokinetics: Drug development's broken stair
There’s a blog post that floats around the Internet about the metaphor of the broken stair.
Grades of quality of Panama hats
Grades of Quality of Panama Hats by Brent Black, the world's leading designer of fine quality Montecristi Panaman Hats.
MTIA v1: Meta’s first-generation AI inference accelerator
In 2020, we initiated the Meta Training and Inference Accelerator (MTIA) family of chips to support our evolving AI workloads, starting with an inference accelerator ASIC for deep learning recommendation models (DLRMs).
Mars will have a lot of wicker furniture
So we've landed on Mars, now we need somewhere to sit. It's pretty expensive to ship furniture from earth, so time to start building some chairs.
Top-Down LR Parsing
This semester, I am teaching Compilers , and I'm quite happy with how I covered parsing this year. Specifically, I taught students to build a top-down LR parser and it seemed to go relatively well.
Abusing vector search for texts, maps, and chess
Vector search isn't just for AI embeddings. Use HNSW for geo-spatial queries, stock covariance, chess positions, text tokens, and multi-modal recommendations with custom metrics.
How to add documentation to your product life cycle
An evaluation of different processes used to create documentation, and their advantages and disadvantages for the organization, the customer, and the writer.
Continuous batching to increase LLM inference throughput and reduce p50 latency
In this blog, we discuss continuous batching, a critical systems-level optimization that improves both throughput and latency under load for LLMs.