GoKawiil - Latest Tech News & Aggregated Headlines

Language models pack billions of concepts into 12k dimensions

news.ycombinator.com Unknown 2025-09-25 23:54:20

In a recent 3Blue1Brown video series on transformer models, Grant Sanderson posed a fascinating question: How can a relatively modest embedding space of 12,288 dimensions (GPT-3) accommodate millions of distinct real-world concepts? The answer lies at the intersection of high-dimensional geometry and a remarkable mathematical result known as the Johnson-Lindenstrauss lemma. While exploring this question, I discovered something unexpected that led to an interesting collaboration with Grant and a

Topics: dimensional embedding high space vectors

Shop Amazon

Language Models Pack Billions of Concepts into 12k Dimensions

news.ycombinator.com Unknown 2025-09-27 10:54:20

In a recent 3Blue1Brown video series on transformer models, Grant Sanderson posed a fascinating question: How can a relatively modest embedding space of 12,288 dimensions (GPT-3) accommodate millions of distinct real-world concepts? The answer lies at the intersection of high-dimensional geometry and a remarkable mathematical result known as the Johnson-Lindenstrauss lemma. While exploring this question, I discovered something unexpected that led to an interesting collaboration with Grant and a

Topics: dimensional embedding high space vectors

Shop Amazon

Language Models Pack Billions of Concepts into 12,000 Dimensions

news.ycombinator.com Unknown 2025-09-27 15:54:20

In a recent 3Blue1Brown video series on transformer models, Grant Sanderson posed a fascinating question: How can a relatively modest embedding space of 12,288 dimensions (GPT-3) accommodate millions of distinct real-world concepts? The answer lies at the intersection of high-dimensional geometry and a remarkable mathematical result known as the Johnson-Lindenstrauss lemma. While exploring this question, I discovered something unexpected that led to an interesting collaboration with Grant and a

Topics: dimensional embedding high space vectors

Shop Amazon

Vector database that can index 1B vectors in 48M

news.ycombinator.com Talip Ozturk 2025-09-28 12:56:18

We are excited to announce Vectroid, a serverless vector search solution that delivers exceptional accuracy and low latency in a cost effective package. Vectroid is not just another vector search solution—it’s a search engine that performs and scales in all scenarios. Why we built Vectroid Talk to any team working with large, low latency vector workloads and you’ll hear a familiar story: something always has to give. Vector databases often make significant tradeoffs between speed, accuracy, an

Topics: cost latency search vector vectroid

Shop Amazon

Will Amazon S3 Vectors kill vector databases or save them?

news.ycombinator.com Unknown 2025-10-07 04:35:46

Not too long ago, AWS dropped something new: S3 Vectors. It’s their first attempt at a vector storage solution, letting you store and query vector embeddings for semantic search right inside Amazon S3. At a glance, it looks like a lightweight vector database running on top of low-cost object storage—at a price point that is clearly attractive compared to many dedicated vector database solutions. amazon s3 vectors.png Naturally, this sparked a lot of hot takes. I’ve seen folks on social media

Topics: data s3 storage vector vectors

Shop Amazon

Will Amazon S3 Vectors Kill Vector Databases–Or Save Them?

news.ycombinator.com Unknown 2025-10-08 00:35:46

Not too long ago, AWS dropped something new: S3 Vectors. It’s their first attempt at a vector storage solution, letting you store and query vector embeddings for semantic search right inside Amazon S3. At a glance, it looks like a lightweight vector database running on top of low-cost object storage—at a price point that is clearly attractive compared to many dedicated vector database solutions. amazon s3 vectors.png Naturally, this sparked a lot of hot takes. I’ve seen folks on social media

Topics: data s3 storage vector vectors

Shop Amazon

The maths you need to start understanding LLMs

news.ycombinator.com Unknown 2025-10-05 06:10:47

The maths you need to start understanding LLMs Actually coming up with ideas like GPT-based LLMs and doing serious AI research requires serious maths. But the good news is that if you just want to understand how they work, while it does require some maths, if you studied it at high-school at any time since the 1960s, you did all of the groundwork then: vectors, matrices, and so on. One thing to note -- what I'm covering here is what you need to know to understand inference -- that is, using an

Topics: dimensional matrix points space vector

Shop Amazon

The repercussions of a typo in C++ & Rust

news.ycombinator.com Unknown 2025-10-16 16:22:01

The repercussions of missing an Ampersand in C++ & Rust Copying vs Passing by reference TL;DR There’s a funny typo that causes someone to copy data instead of “referencing” in C++. Rust is nice because it provides defaults that protect you from some of these “dumb” mistakes. In this example, I’ll go over how the “move by default” can prevent us from introducing this subtle behavior. Motivation I originally hesitated to write this because I thought the topic was too “obvious”, but I did it a

Topics: copy data request rust vec

Shop Amazon

Implementing a Foil Sticker Effect

news.ycombinator.com Nikolaos Papadopoulos 2025-10-18 20:47:21

In this post, I’ll walk you through how to create a custom shader in Three.js that simulates the look of a foil sticker, complete with angle-dependent iridescence and sparkling metallic flakes. The goal is to capture that premium, holographic effect you see on collectible stickers, trading cards, and high-end packaging, but to render it in real time directly in the browser. Iridescence If you’ve ever tilted a holographic sticker or watched sunlight catch on a soap bubble, you’ve seen iridescen

Topics: float peel uniform vec3 vertex

Shop Amazon

From multi-head to latent attention: The evolution of attention mechanisms

news.ycombinator.com Unknown 2025-10-19 10:45:24

From Multi-Head to Latent Attention: The Evolution of Attention Mechanisms Vinithavn 7 min read · 15 hours ago 15 hours ago -- Listen Share Press enter or click to view image in full size What is attention? In any autoregressive model, the prediction of the future tokens is based on some preceding context. However, not all the tokens within this context equally contribute to the prediction, because some tokens might be more relevant than others. The attention mechanism addresses this by allow

Topics: attention context key query vectors

Shop Amazon

From Multi-Head to Latent Attention: The Evolution of Attention Mechanisms

news.ycombinator.com Unknown 2025-10-19 20:45:24

From Multi-Head to Latent Attention: The Evolution of Attention Mechanisms Vinithavn 7 min read · 15 hours ago 15 hours ago -- Listen Share Press enter or click to view image in full size What is attention? In any autoregressive model, the prediction of the future tokens is based on some preceding context. However, not all the tokens within this context equally contribute to the prediction, because some tokens might be more relevant than others. The attention mechanism addresses this by allow

Topics: attention context key query vectors

Shop Amazon

Macintosh Drawing Software Compared

news.ycombinator.com Unknown 2025-11-11 21:09:17

Below is a list of various drawing applications that are available to use with System 7 (68K). I’m mostly interested in apps that can do both pixel and vector, and that support layers and transparency. The table below might be useful for at-a-glance comparisons. Apps run in both mono or colour, unless stated otherwise. Deneba artWORKS can do both pixel and vector, has layers and partial transparency. The bottom layer is opaque and all layers on top are transparent. Collections of pixels are ma

Topics: freehand layers objects pixel vector

Shop Amazon

Going faster than memcpy

news.ycombinator.com Dheeraj R Reddy 2025-11-27 02:59:03

Going faster than memcpy While profiling Shadesmar a couple of weeks ago, I noticed that for large binary unserialized messages (>512kB) most of the execution time is spent doing copying the message (using memcpy ) between process memory to shared memory and back. I had a few hours to kill last weekend, and I tried to implement a faster way to do memory copies. Autopsy of memcpy Here’s the dumb of perf when running pub-sub for messages of sizes between 512kB and 2MB. Children Self Shared Ob

Topics: const loop size_t svec void

Shop Amazon

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality

venturebeat.com Ben Dickson 2025-12-02 17:11:20

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now A new study from the Anthropic Fellows Program reveals a technique to identify, monitor and control character traits in large language models (LLMs). The findings show that models can develop undesirable personalities (e.g., becoming malicious, excessively agreeable, or prone to making things up) either in response to user prompts or as an

Topics: ai model persona training vectors

Shop Amazon

Google’s AI coding agent Jules is now out of beta

techcrunch.com Jagmeet Singh 2025-12-03 22:00:00

Google on Wednesday launched its AI coding agent, Jules, out of beta, just over two months after its public preview debut in May. Powered by Gemini 2.5 Pro, Jules is an asynchronous, agent-based coding tool that integrates with GitHub, clones codebases into Google Cloud virtual machines, and uses AI to fix or update code while developers focus on other tasks. Google initially announced Jules as a Google Labs project in December and made it available to beta testers through a public preview at

Topics: beta google jules korevec said

Shop Amazon

Anthropic wants to stop AI models from turning evil - here's how

zdnet.com Radhika Rajkumar 2025-12-08 22:30:28

Lyudmila Lucienne/Getty ZDNET's key takeaways New research from Anthropic identifies model characteristics, called persona vectors. This helps catch bad behavior without impacting performance. Still, developers don't know enough about why models hallucinate and behave in evil ways. Why do models hallucinate, make violent suggestions, or overly agree with users? Generally, researchers don't really know. But Anthropic just found new insights that could help stop this behavior before it happen

Topics: anthropic model models persona vectors

Shop Amazon

Persona vectors: Monitoring and controlling character traits in language models

news.ycombinator.com Unknown 2025-12-10 08:38:06

Language models are strange beasts. In many ways they appear to have human-like “personalities” and “moods,” but these traits are highly fluid and liable to change unexpectedly. Sometimes these changes are dramatic. In 2023, Microsoft's Bing chatbot famously adopted an alter-ego called "Sydney,” which declared love for users and made threats of blackmail. More recently, xAI’s Grok chatbot would for a brief period sometimes identify as “MechaHitler” and make antisemitic comments. Other personali

Topics: model persona training trait vectors

Shop Amazon

AV-Racer Devlog (1): Getting a functional car model

news.ycombinator.com Unknown 2025-12-19 17:23:09

wassimulator; I make stuff Next article ► AV-Racer Devlog (1): Getting a functional car model est. reading time: 10 minutes The first question revolving around the interest of making a car game was figuring out the methods used by car games in simulating a car body. My modest lacunary conglamurate of car facts was all I had to start with, the general idea was to simulate car and tire forces that would move the thing through the might of physics. The right approach to this problem became a matt

Topics: angular car forces vector velocity

Shop Amazon

Test Results for AMD Zen 5

news.ycombinator.com Unknown 2025-12-23 20:44:57

Post by agner » 2025-07-26, 12:43:13 I have now finished testing the Zen 5. Thank you to the people who have helped running test scripts for me.My test results for the AMD Zen 5 are impressive. It has a lot of features that increase different aspects of the CPU performance to new levels, never seen before.Most importantly, the instruction fetch rate is increased from 16 to 32 bytes per clock cycle. The 16-bytes fetch rate has been a serious bottleneck in both Intel and AMD processors through ma

Topics: cycle instruction instructions vector zen

Shop Amazon

Generic Containers in C: Vec

news.ycombinator.com Unknown 2025-12-20 10:04:26

Generic Containers in C: vec Martin Uecker, 2025-07-20 I discuss the implementation of type and bounds safe generic containers in C. Previously, I discussed a span type, and bounds checking using arrays. Here, I will discuss a vector type. A vector type is essentially a resizable array. A vector type could be used as in the following example. int main() { vec(int) *vec_ptr = calloc(1, sizeof *vec_ptr); if (!vec_ptr) // memory out abort(); for (int i = 0; i < 10; i++) vec_push(int, &vec

Topics: int type vec vec_ptr vector

Shop Amazon

This industrial AI startup is winning over customers by saying it won’t get acquired

techcrunch.com Sean O'Kane 2025-12-27 22:37:47

When industrial AI startup CVector meets with manufacturers, utility providers, and other prospective customers, the founders are often asked the same question: will you still be here in six months? A year? It’s a fair concern in an environment where the biggest, richest tech companies are luring top talent with eye-watering salaries and increasingly targeting rising AI startups with elaborate acquihire deals. The answer that CVector founders Richard Zhang and Tyler Ruggles give every time is

Topics: customers cvector ruggles said zhang

Shop Amazon

Vector Tiles are deployed on OpenStreetMap.org

news.ycombinator.com Unknown 2025-12-28 22:20:16

We are happy to announce the deployment of Vector Tiles on OpenStreetMap Foundation servers and the publication of the layer on the OSM website! We have been working hard to bring you a fresh look to OSM data, paired with exciting technological upgrades. Work has been progressing since last year. In June 2024 we shared progress, including the launch of the vector tiles demo site, as well as details on the technical background on the tools being used. Since then we have put the tile generation p

Topics: new styles tile tiles vector

Shop Amazon

Using uninitialized memory for fun and profit (2008)

news.ycombinator.com Russ Cox 2025-12-26 03:11:09

Using Uninitialized Memory for Fun and Profit Posted on Friday, March 14, 2008. This is the story of a clever trick that's been around for at least 35 years, in which array values can be left uninitialized and then read during normal operations, yet the code behaves correctly no matter what garbage is sitting in the array. Like the best programming tricks, this one is the right tool for the job in certain situations. The sleaziness of uninitialized data access is offset by performance improveme

Topics: dense set sparse time vector

Shop Amazon

Linda Hamilton rocks Stranger Things’ S5 extended teaser

arstechnica.com Unknown 2026-01-12 21:06:45

Netflix has finally released an extended teaser for the fifth and final season of Stranger Things, airing late this year. It's got everything we could hope for in terms of the conclusive showdown—spooky shots of the Upside Down bleeding into Hawkins, attacking demo-dogs, flamethrowers—plus an armed and dangerous Linda Hamilton taking on the monsters beside our plucky Hawkins crew. (Spoilers for prior seasons below.) S4 ended with Vecna—the Big Bad behind it all—opening the gate that allowed th

Topics: anniversary disappearance fifth hawkins vecna

Shop Amazon

The messy reality of SIMD (vector) functions

news.ycombinator.com Ivica Bogosavljević 2026-02-02 12:14:52

We’ve discussed SIMD and vectorization extensively on this blog, and it was only a matter of time before SIMD (or vector) functions came up. In this post, we explore what SIMD functions are, when they are useful, and how to declare and use them effectively. A SIMD function is a function that processes more than one piece of data. Take for example a mathematical sin function: double sin(double angle); This function takes one double and returns one double. The vector version that processes four

Topics: compiler double function simd vector

Shop Amazon

Muvera: Making multi-vector retrieval as fast as single-vector search

news.ycombinator.com Unknown 2026-02-21 04:29:34

Neural embedding models have become a cornerstone of modern information retrieval (IR). Given a query from a user (e.g., “How tall is Mt Everest?”), the goal of IR is to find information relevant to the query from a very large collection of data (e.g., the billions of documents, images, or videos on the Web). Embedding models transform each datapoint into a single-vector “embedding”, such that semantically similar datapoints are transformed into mathematically similar vectors. The embeddings are

Topics: embedding multi retrieval similarity vector

Shop Amazon

Neural Texture Compression demo shows it can do wonders for VRAM usage

techspot.com Alfonso Maruccia 2026-03-05 22:35:00

Serving tech enthusiasts for over 25 years.TechSpot means tech analysis and advice you can trust In context: Modern game engines can put severe strain on today's hardware. However, Nvidia's business decisions have left many GPUs with less VRAM than they should. Fortunately, improved texture compression in games helps make the most of what's available. Neural Texture Compression (NTC) is a new technique that improves texture quality while reducing VRAM usage. It relies on a specialized neural n

Topics: cooperative intel ntc texture vectors

Shop Amazon

Painting with Math: A Gentle Study of Raymarching (2023)

news.ycombinator.com Unknown 2026-03-04 02:52:17

Most of my experience writing GLSL so far focused on enhancing pre-existing Three.js/React Three Fiber scenes that contain diverse geometries and materials with effects that wouldn't be achievable without shaders, such as my work with dispersion and particle effects. However, during my studies of shaders, I always found my way to Shadertoy, which contains a multitude of impressive 3D scenes featuring landscapes, clouds, fractals, and so much more, entirely implemented in GLSL. No geometries. No

Topics: distance float raymarching scene vec3

Shop Amazon

Painting with Math: A Gentle Study of Raymarching

news.ycombinator.com Unknown 2026-03-05 14:52:17

Most of my experience writing GLSL so far focused on enhancing pre-existing Three.js/React Three Fiber scenes that contain diverse geometries and materials with effects that wouldn't be achievable without shaders, such as my work with dispersion and particle effects. However, during my studies of shaders, I always found my way to Shadertoy, which contains a multitude of impressive 3D scenes featuring landscapes, clouds, fractals, and so much more, entirely implemented in GLSL. No geometries. No

Topics: distance float raymarching scene vec3

Shop Amazon

3D printable 6" f/5 compact travel telescope model

news.ycombinator.com Unknown 2026-03-04 08:23:05

README ON MY PAGE: https://la3emedim.fr/work/nested/smallest/ ATTENTION : Ne dirigez jamais le télescope Smallest vers le Soleil — c’est extrêmement dangereux pour vos yeux. WARNING : Never point the Smallest telescope at the Sun — it's extremely dangerous for your eyes. ENGLISH: THE SMALLEST Telescope kit I’m making the project files freely available. I’ve carefully designed them to be accessible and easy to use — but if you’re looking for a flawless finish, tested assembly, and a 2-year

Topics: avec la le pour print

Shop Amazon

Latest Tech News

Language models pack billions of concepts into 12k dimensions

Language Models Pack Billions of Concepts into 12k Dimensions

Language Models Pack Billions of Concepts into 12,000 Dimensions

Vector database that can index 1B vectors in 48M

Will Amazon S3 Vectors kill vector databases or save them?

Will Amazon S3 Vectors Kill Vector Databases–Or Save Them?

The maths you need to start understanding LLMs

The repercussions of a typo in C++ & Rust

Implementing a Foil Sticker Effect

From multi-head to latent attention: The evolution of attention mechanisms

From Multi-Head to Latent Attention: The Evolution of Attention Mechanisms

Macintosh Drawing Software Compared

Going faster than memcpy

New ‘persona vectors’ from Anthropic let you decode and direct an LLM’s personality

Google’s AI coding agent Jules is now out of beta

Anthropic wants to stop AI models from turning evil - here's how

Persona vectors: Monitoring and controlling character traits in language models

AV-Racer Devlog (1): Getting a functional car model

Test Results for AMD Zen 5

Generic Containers in C: Vec

This industrial AI startup is winning over customers by saying it won’t get acquired

Vector Tiles are deployed on OpenStreetMap.org

Using uninitialized memory for fun and profit (2008)

Linda Hamilton rocks Stranger Things’ S5 extended teaser

The messy reality of SIMD (vector) functions

Muvera: Making multi-vector retrieval as fast as single-vector search

Neural Texture Compression demo shows it can do wonders for VRAM usage

Painting with Math: A Gentle Study of Raymarching (2023)

Painting with Math: A Gentle Study of Raymarching

3D printable 6" f/5 compact travel telescope model

About GoKawiil

Privacy

Advertising

Latest Tech News

Trending Topics

Hot Now

Popular

Emerging

About GoKawiil

Privacy

Advertising