Latest Tech News

Stay updated with the latest in technology, AI, cybersecurity, and more

Filtered by: mod Clear Filter

Experimenting with Local LLMs on macOS

So, this blog post will be about LLMs, and everyone has opinions about that. To be upfront about it, I’m a skeptic (bordering on hater), yet I like experimenting with stuff so I download and run them locally on my Mac. And I’ll teach you how to do it too, if you’d like! Some call them fancy autocomplete, some argue that they are sentient and should have rights. The truth is somewhere in between. Yes, they perform next word prediction, but it’s so complex that there’s nontrivial emergent behavio

Bose QuietComfort Ultra headphones will get a 2nd-gen update next month

The popular Bose QuietComfort Ultra headphones already got one update back in the summer, but the company has now announced a bigger deserving of a 2nd-generation label … Back in June, the company jumped on the AI bandwagon with what it described as enhanced adaptive noise cancellation. It also claimed better call quality and wireless charging capabilities. Those improvements weren’t enough to see it become an official second-generation model, but Engadget reports that’s what the company will

If these iPhone 17 Air rumors are confirmed on Tuesday, I'm saying goodbye to my phone

The iPhone Plus model (pictured) may potentially be replaced by the iPhone Air/Slim. Kerry Wan/ZDNET Follow ZDNET: Add us as a preferred source on Google. ZDNET's key takeaways iPhone 17 Air may debut as Apple's thinnest phone ever. Single rear camera shows Apple's thinness trade-offs. Expected to debut on Sept. 9, 2025, priced around $900. Apple is rumored to be spicing things for this Tuesday's iPhone event. It could introduce an ultra-thin model for the 2025 iPhone lineup called the iPh

Google to make it easier to access AI Mode as default

Google plans to make it easier for users to access AI mode by allowing them to set it as the default, replacing the traditional blue links. AI mode is an advanced version of Google Search that uses large language models to summarise information from the web, so you can spend more time on Google than visiting websites. Google AI mode advanced analysis Source: BleepingComputer Google AI mode can answer complex answers, process images, summarize information on the web, create tables, graphs, ch

Using Claude Code to modernize a 25-year-old kernel driver

As a bit of background, one of my hobbies is helping people recover data from old tape cartridges, such as QIC-80 tapes, which were a rather popular backup medium in the 1990s among individuals, small businesses, BBS operators, and the like. I have a soft spot for tape media; there’s something about the tactile sensation of holding these tapes in my hands that makes the whole process very joyful, even though QIC tapes are notorious for their many design flaws. With some careful inspection and re

Are bad incentives to blame for AI hallucinations?

A new research paper from OpenAI asks why large language models like GPT-5 and chatbots like ChatGPT still hallucinate, and whether anything can be done to reduce those hallucinations. In a blog post summarizing the paper, OpenAI defines hallucinations as “plausible but false statements generated by language models,” and it acknowledges that despite improvements, hallucinations “remain a fundamental challenge for all large language models” — one that will never be completely eliminated. To ill

I Tried Gemini's 'Nano Bananas' for Image Editing. The AI Slipups Were Obvious

After seeing all the banana-fanfare for Google's newest generative AI tool, I knew I had to take it for a spin. Named Gemini 2.5 Flash Image, the model upgrades your ability to edit your photos natively in Gemini. AI enthusiasts have referred to it as the "nano bananas" model, spurred on by a series of banana-themed teasers from Google execs. In the few weeks it's been out, people have created over 200 million AI images, and over 10 million people have signed up to use the Gemini app, according

The key to getting MVC correct is understanding what models are

Smalltalk MVC is defined in Design Pattern as: MVC Consists of three kinds of objects. The Model is the application object, the View is its screen presentation, and the Controller defines the way the user interface reacts to user input. However this definition has been abused over the years - Back in 2003 I gave a talk citing how bad Apple’s definition was. At the time it stated: A view object knows how to display and possibly edit data from the application’s model… A controller object acts a

Over 80% of sunscreen performed below their labelled efficacy (2020)

The use of effective sunscreen can reduce the harm caused to the skin by ultraviolet rays (UV) and slow down skin aging. The Consumer Council tested 30 models of sunscreen for daily use and over 80% of them were found to perform below their respective labelled efficacy. The measured sunscreen efficacy of 4 models were below SPF15, of which 2 were sunscreen products with very high protection i.e. labelled with SPF50+. Among the 23 models using the “PA System” which is commonly adopted by Asian co

The MVC definition has been abused

Smalltalk MVC is defined in Design Pattern as: MVC Consists of three kinds of objects. The Model is the application object, the View is its screen presentation, and the Controller defines the way the user interface reacts to user input. However this definition has been abused over the years - Back in 2003 I gave a talk citing how bad Apple’s definition was. At the time it stated: A view object knows how to display and possibly edit data from the application’s model… A controller object acts a

AYANEO just undercut the Odin 3 for the cheapest 8 Elite handheld, and you can pre-order today

AYANEO TL;DR The KONKR Pocket FIT and FIT Elite are available for pre-order now, with early bird discounts. Pricing on the G3 Gen 3 model starts at $239, with the 8 Elite model starting at $269. The AYN Odin 3, which also has an 8 Elite and launches tomorrow, starts at $299. They say competition is good for the consumer, and never has that been felt more by the Android gaming handheld community than in the past few weeks. First, AYANEO and AYN released competing dual-screen clamshells in the

AI's not 'reasoning' at all - how this team debunked the industry hype

Pulse/Corbis via Getty Images Follow ZDNET: Add us as a preferred source on Google. ZDNET's key takeaways We don't entirely know how AI works, so we ascribe magical powers to it. Claims that Gen AI can reason are a "brittle mirage." We should always be specific about what AI is doing and avoid hyperbole. Ever since artificial intelligence programs began impressing the general public, AI scholars have been making claims for the technology's deeper significance, even asserting the prospect

GLM 4.5 with Claude Code

GLM Coding Plan — designed for Claude Code users, starting at $3/month to enjoy a premium coding experience! GLM-4.5 and GLM-4.5-Air are our latest flagship models, purpose-built as foundational models for agent-oriented applications. Both leverage a Mixture-of-Experts (MoE) architecture. GLM-4.5 has a total parameter count of 355B with 32B active parameters per forward pass, while GLM-4.5-Air adopts a more streamlined design with 106B total parameters and 12B active parameters. Both models sh

Two authors accuse Apple of illegally training AI models on pirated books

A new proposed class action suit was filed in the federal court in Northern California today, accusing Apple of illegally using books to train its AI models. Here are the details. Authors base the accusation on Apple’s own documents As reported by Reuters, authors Grady Hendrix and Jennifer Robertson are accusing Apple of using a pirated dataset, in which their work was included. From the lawsuit: “But Apple is building part of this new enterprise using Books3, a dataset of pirated copyrighte

OpenAI reorganizes research team behind ChatGPT’s personality

OpenAI is reorganizing its Model Behavior team, a small but influential group of researchers who shape how the company’s AI models interact with people, TechCrunch has learned. In an August memo to staff seen by TechCrunch, OpenAI’s chief research officer Mark Chen said the Model Behavior team — which consists of roughly 14 researchers — would be joining the Post Training team, a larger research group responsible for improving the company’s AI models after their initial pre-training. As part o

What Are AI Hallucinations? Why Chatbots Make Things Up, and What You Need to Know

If you've used ChatGPT, Google Gemini, Grok, Claude, Perplexity or any other generative AI tool, you've probably seen them make things up with complete confidence. This is called an AI hallucination -- although one research paper suggests we call it BS instead -- and it's an inherent flaw that should give us all pause when using AI. Hallucinations happen when AI models generate information that looks plausible but is false, misleading or entirely fabricated. It can be as small as a wrong date i

Anthropic Agrees to $1.5 Billion Settlement for Downloading Pirated Books to Train AI

Anthropic has agreed to pay $1.5 billion to settle a lawsuit brought by authors and publishers over its use of millions of copyrighted books to train the models for its AI chatbot Claude, according to a legal filing posted online. A federal judge found in June that Anthropic’s use of 7 million pirated books was protected under fair use but that holding the digital works in a “central library” violated copyright law. The judge ruled that executives at the company knew they were downloading pirat

Apertus 70B: Truly Open - Swiss LLM by ETH, EPFL and CSCS

Apertus Table of Contents Model Summary Apertus is a 70B and 8B parameter language model designed to push the boundaries of fully-open multilingual and transparent models. The model supports over 1000 languages and long context, it uses only fully compliant and open training data, and achieves comparable performance to models trained behind closed doors. The model is a decoder-only transformer, pretrained on 15T tokens with a staged curriculum of web, code and math data. The model uses a new

Relace (YC W23) Is Hiring for Code LLMs (SF)

Hey, we're a highly technical team building code generation models, and growing fast. We're looking for people who are down to scrap and love to build -- on both technical and GTM/Devrel roles. If you have a Physics, Math, CS degree; and training fast codegen models is something that piques your interest, please email me directly at [email protected].

Data modeling guide for real-time analytics with ClickHouse

This article was written as part of my services Querying billions of weather records and getting results in under 200 milliseconds isn’t theory; it’s what real-time analytics solutions provide. Processing streaming IoT data from thousands of sensors while delivering real-time dashboards with no lag is what certain business domains need. That’s what you’ll learn at the end of this guide through building a ClickHouse-modeled analytics use case. You’ll learn how to land data in ClickHouse that is

How big are our embeddings now and why?

Sep 1 2025 #embeddings #openai #anthropic #huggingface #dimensionality A few years ago, I wrote a paper on embeddings. At the time, I wrote that 200-300 dimension embeddings were fairly common in industry, and that adding more dimensions during training would create diminishing returns for the effectiveness of your downstream tasks (classification, recommendation, semantic search, topic modeling, etc.) I wrote the paper to be resilient to changes in the industry since it focuses on fundamenta

Gemini's 'Nano Bananas' AI Image Editing Is Fun, but I Ran Into Too Many Slipups

I've been trying out the image editing capabilities in Google's latest generative AI model, and you can, too. Named Gemini 2.5 Flash Image, it upgrades the ability to edit your photos natively in Gemini. AI enthusiasts have referred to it as the "nano bananas" model, spurred on by a series of teasers from Google execs. Google has been investing heavily in its generative media models this year, dropping updated versions of its image and video generator models at its annual I/O developers confere

Data Modeling Guide for Real-Time Analytics with ClickHouse

This article was written as part of my services Querying billions of weather records and getting results in under 200 milliseconds isn’t theory; it’s what real-time analytics solutions provide. Processing streaming IoT data from thousands of sensors while delivering real-time dashboards with no lag is what certain business domains need. That’s what you’ll learn at the end of this guide through building a ClickHouse-modeled analytics use case. You’ll learn how to land data in ClickHouse that is

ChatGPT Glossary: 56 AI Terms Everyone Should Know

AI is rapidly changing the world around us. It's eliminating jobs and flooding the internet with slop. Thanks to the massive popularity of ChatGPT to Google cramming AI summaries at the top of its search results, AI is completely taking over the internet. With AI, you can get instant answers to pretty much any question. It can feel like talking to someone who has a doctoral degree in everything. But that aspect of AI chatbots is only one part of the AI landscape. Sure, having ChatGPT help do yo

Relace (YC W23) Is Hiring for Code LLM's (SF)

Hey, we're a highly technical team building code generation models, and growing fast. We're looking for people who are down to scrap and love to build -- on both technical and GTM/Devrel roles. If you have a Physics, Math, CS degree; and training fast codegen models is something that piques your interest, please email me directly at [email protected].

Lava RGB

Although the power module includes a Saturn-style DIN connector, I don’t own a compatible cable, and prefer to use VGA cables. So for my install, I decided to add a SNES multiout port. Compared to the NESRGB, Lava RGB 2.0 is definitely a worthy contender. The palette switching OSD, and ability to easily update firmware, are improvements, and I wonder if future firmware updates might add more OSD-based options. However, one thing Lava RGB does not do is process and output audio like the NESRGB.

Microsoft Goes Back to BASIC, Open-Sources Bill Gates’ Code

In the era of vibe coding, when even professionals are pawning off their programming work on AI tools, Microsoft is throwing it all the way back to the language that launched a billion devices. On Wednesday, the company announced that it would make the source code for Microsoft BASIC for the 6502 Version 1.1 publicly available and open-source. The code is now uploaded to GitHub under an MIT license (with a cheeky commit time stamp of “48 years ago”). Microsoft called the code—written by the com

All the Top New Gadgets We Saw at IFA Berlin 2025

Every September, hundreds of thousands of visitors descend upon Berlin for IFA, one of the largest trade tech conferences in Europe. IFA heralds the unofficial start of Techtember, when companies launch products in anticipation of the holiday shopping season. For over 100 years, companies, financiers, and journalists have crossed the Spree and eaten cold liver sausage for breakfast, just to check out the latest innovations in consumer electronics. This year, those attendees include me and Gear

Bose announces QuietComfort Ultra headphones with support for lossless audio over USB-C

is a senior reporter who’s been covering and reviewing the latest gadgets and tech since 2006, but has loved all things electronic since he was a kid. Posts from this author will be added to your daily email digest and your homepage feed. Bose has announced a second-generation version of its QuietComfort Ultra wireless headphones it first introduced nearly two years ago. The name is the same, but the new Ultra finally adds support for listening to lossless audio over a USB-C cable, while still

The Pretty Screen on Lenovo’s Legion Go 2 Will Almost Make You Forget About Its Price

Why change what works? In the case of Lenovo’s Legion Go 2, the question morphs into “Why change what makes you distinct?” Lenovo’s upcoming sequel to its Legion Go handheld gaming PC is more powerful, has grips made for human hands, and sports a prettier screen. Its blood is flowing with the same DNA as Lenovo’s odd handheld, but now it comes with a price tag designed to make your wallet weep. Here’s the kicker. At IFA 2025, Lenovo declared the Legion Go 2 will start at $1,050 for a version we