Latest Tech News

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

Filtered by: ls Clear Filter

GPT-OSS vs. Qwen3 and a detailed look how things evolved since GPT-2

OpenAI just released their new open-weight LLMs this week: gpt-oss-120b and gpt-oss-20b, their first open-weight models since GPT-2 in 2019. And yes, thanks to some clever optimizations, they can run locally (but more about this later). This is the first time since GPT-2 that OpenAI has shared a large, fully open-weight model. Earlier GPT models showed how the transformer architecture scales. The 2022 ChatGPT release then made these models mainstream by demonstrating concrete usefulness for wri

AI’s promise of opportunity masks a reality of managed displacement

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 Cognitive migration is underway. The station is crowded. Some have boarded while others hesitate, unsure whether the destination justifies the departure. Future of work expert and Harvard University Professor Christopher Stanton commented recently that the uptake of AI has been tremendous and observed that it is an “extraordinarily fast-di

A large number of protocols on Ethereum and Solana blockchains have no revenue

Have you heard of disguised unemployment? It refers to a situation where a portion of the workforce appears to be employed, but isn't contributing to the economy's output. Consider the massive capital expenditure loss from ghost cities, which represent unoccupied infrastructure. Something similar can be said for the top smart contract blockchains, which hosts hundreds of decentralized protocols. Of these, only a minority are generating revenue, while the rest produce no yield, loosely represent

ECScape: Understanding IAM Privilege Boundaries in Amazon ECS

This post is Part 2 of our educational series on Amazon ECS security. In Part 1 – Under the Hood of Amazon ECS on EC2, we explored how the ECS agent, IAM roles and the ECS control plane provide credentials to tasks. Here we’ll demonstrate how those mechanisms can lead to a known risk when tasks with different privilege levels share the same EC2 host. This cross-task credential exposure highlights the inherent risks of relying on per-task IAM scoping and task execution boundaries when workloads s

OpenAI Brings Back Fan-Favorite GPT-4o After a Massive User Revolt

After a disastrous 72 hours that saw its most loyal users in open revolt, OpenAI is making a major U-turn. In a series of posts on X (formerly Twitter) Sunday, CEO Sam Altman announced that the company is bringing back its beloved older AI models, including GPT-4o, and dramatically increasing usage limits for paying subscribers, a clear peace offering to a furious customer base. The move comes just days after the botched rollout of GPT-5, the company’s latest and most powerful model. The launc

How to Use Parallels to Run Windows on a Mac

While Apple is very keen that its macOS software is only ever run on Macs, there are numerous options for doing the opposite and running Windows on Apple computers. If you're a developer, or need access to both Windows and Mac apps, or simply want the flexibility of two operating systems, this isn't too difficult to set up with the right tools. The option we'll cover here is Parallels Desktop for Mac. This is certainly not the cheapest of your options, but it is fully authorized and legal, as w

My Google Keep notes were a mess — here’s how I got them under control

Joe Maring / Android Authority Google Keep has been my go-to note-taking app for years, but I recently noticed that my disorganized notes were becoming overwhelming, making it harder to find what I was looking for. I had previously tried color-coding notes and pinning important ones, however these fixes were no longer enough. I decided to finally get control over my notes by using features I’d often ignored, including labels and deleting old notes I no longer needed. But I also moved some of m

LAST DAY: Massive Apple deal event – M3 iPad Air all-time low, AirPods, M4 MacBooks $300 off, more [Updated]

As we approach week’s end, we are starting to track a massive selection of Apple deals across its current-generation lineup. From the iPad Air, mini 7, and Pro, right through the the M4 MacBook Air, a series of accessories, and much more, we are looking at some of the best prices of the year and even some new all-time lows that undercut what we saw on Prime Day. Scope it all out down below. Some of today’s deals were already live at Amazon previously, but along with the new shopping event now l

Topics: air deals ipad m4 pro

iPhone Users, Find Calls Faster With This Trick

It can be frustrating to scroll through your iPhone's Recent calls tab to find the right voicemail or to remember when you talked to someone last. But when Apple introduced iOS 18 in 2024, it included a trick that let you find calls faster in your recent call history. That update brought a lot of new features, like customizable home screens and RCS messaging, as well as a search bar in your iPhone's Phone app. This lets you easily search your call history and voicemails. Before iOS 18, your Pho

How to Watch Chelsea vs. AC Milan From Anywhere for Free: Stream Preseason Friendly Soccer

Chelsea's final warm up match before the start of the Premier League season sees the team host AC Milan today at Stamford Bridge. Below, we'll outline the best live TV streaming services to watch every match of the tournament as it happens, wherever you are in the world. We'll also explain how to use a VPN if the match isn't available where you are, along with a full match list. The recent FIFA Club World Cup winners have had a truncated preseason, thanks to their involvement in that tournamen

How To Use Parallels to Run Windows on a Mac

While Apple is very keen that its macOS software is only ever run on Macs, there are numerous options for doing the opposite and running Windows on Apple computers. If you're a developer, or need access to both Windows and Mac apps, or simply want the flexibility of two operating systems, this isn't too difficult to set up with the right tools. The option we'll cover here is Parallels Desktop for Mac. This is certainly not the cheapest of your options, but it is fully authorized and legal, as w

We found stuff AI is pretty good at

is a senior reporter focusing on wearables, health tech, and more with 13 years of experience. Before coming to The Verge, she worked for Gizmodo and PC Magazine. Tech companies keep telling everyone that this or that AI feature is going to change everything. But when you press them for examples, real, concrete examples of how those AI tools should be used, the answers tend to be lackluster. Sometimes AI tools feel so open-ended, it’s hard to know where to start or what the best way to use them

My Lethal Trifecta talk at the Bay Area AI Security Meetup

In the pirate case there’s no real damage done... but the risks of real damage from prompt injection are constantly increasing as we build more powerful and sensitive systems on top of LLMs. I think this is why we still haven’t seen a successful “digital assistant for your email”, despite enormous demand for this. If we’re going to unleash LLM tools on our email, we need to be very confident that this kind of attack won’t work. My hypothetical digital assistant is called Marvin. What happens i

60 malicious Ruby gems downloaded 275,000 times steal credentials

Sixty malicious Ruby gems containing credential-stealing code have been downloaded over 275,000 times since March 2023, targeting developer accounts. The malicious Ruby gems were discovered by Socket, which reports they targeted primarily South Korean users of automation tools for Instagram, TikTok, Twitter/X, Telegram, Naver, WordPress, and Kakao. RubyGems is the official package manager for the Ruby programming language, enabling the distribution, installation, and management of Ruby librari

How I use Tailscale

The Tailscale logo I’ve been using Tailscale for around four years to connect my disparate devices, servers and apps together. I wanted to talk a bit about how I use it, some cool features you might not know about, and some stumbling blocks I encountered. I’m not sure Tailscale needs an introduction for the likely audience of this blog, but I’ll give one anyway. Tailscale is basically a WireGuard orchestration service, with lots of nice features sprinkled on top. It’s a subscription product, b

The current state of LLM-driven development

I spent the past ~4 weeks trying out all the new and fancy AI tools for software development. Let’s get a few things out of the way: Learning how to use LLMs in a coding workflow is trivial. There is no learning curve. You can safely ignore them if they don’t fit your workflows at the moment. LLMs won’t magically make you deliver production-ready code If you can’t read the code and spot issues, they’re hard to use past the PoC stage They have terrible code organization skills, making them los

Hoto Snapbloq Tools Review: Attractive Power Tools

Tools often have a utilitarian design. Their purpose is to get stuff done, so their shape is exclusively focused on getting stuff done. This is where the Hoto SnapBloq toolkit is unique: These tools emphasize style and design to prove that functionality doesn’t need to have a rugged, utilitarian, and (typically) unappealing style. Hoto’s SnapBloq line is a set of three small power tools meant for tinkerers and hobbyists. The full set consists of a screwdriver, a rotary tool, and a drill, along

Simon Willison's Lethal Trifecta Talk at the Bay Area AI Security Meetup

In the pirate case there’s no real damage done... but the risks of real damage from prompt injection are constantly increasing as we build more powerful and sensitive systems on top of LLMs. I think this is why we still haven’t seen a successful “digital assistant for your email”, despite enormous demand for this. If we’re going to unleash LLM tools on our email, we need to be very confident that this kind of attack won’t work. My hypothetical digital assistant is called Marvin. What happens i

Wild Pigs in California Are Turning Neon Blue on the Inside, Officials Warn

Unexpected exposure to rodenticide is turning California’s wild pigs blue, state authorities say. “I’m not talking about a little blue,” Dan Burton, owner of a wildlife control company in Salinas, California, told The Los Angeles Times. “I’m talking about neon blue, blueberry blue.” Burton was one of the first trappers to discover that local wild pigs had turned blue on the inside. A subsequent investigation by the California Department of Fish and Wildlife (CDFW) found that the pigs had consu

Representing Python notebooks as dataflow graphs

This blog is adapted from our talk at PyCon 2025. marimo is free and open source, available on GitHub. For a free online experience with link sharing, try molab. marimo is a new kind of open-source Python notebook. While traditional notebooks are just REPLs, marimo notebooks are Python programs represented as dataflow graphs. This intermediate representation lets marimo blend the best parts of interactive computing with the reproducibility and reusability of Python software: every marimo notebo

Gear News of the Week: iPhone 17 May Be a Month Away, and Sonos to Raise Prices

If rumors are correct, Apple's annual iPhone event will take place exactly a month from today, on September 9. That's according to a German website citing internal documents from German mobile phone providers, but the date was also previously suggested by Bloomberg's Apple whisperer, Mark Gurman. Leaks about Apple's upcoming smartphone lineup have heated up in recent weeks. Apple is expected to debut four iPhones as usual, with one key distinction. The “Plus” iPhone no longer exists, replaced b

OpenAI Bringing Back More Parasocial Version of ChatGPT After Users Scream and Cry That Their Robot Friend Got Taken Away

ChatGPT's fans have developed intense attachments to some of the "personalities" exhibited by the company's large language models (LLMs) — and as such, many were in a state of panic when the choice to switch between models like GPT-4o, 4.5, and others was abruptly taken away after the release of GPT-5. OpenAI briefly kiboshed the model picker option from ChatGPT as part of the launch of GPT-5, its long-awaited new LLM, which essentially forced everyone to use the latest model whether they liked

Engineer restores pay phones for free public use

An engineer restores pay phones for free public use toggle caption Patrick Schlott Patrick Schlott often finds himself in a cellular dead zone during his drive to work. "You go down the road, you turn the corner and you're behind a mountain and you'll lose cell coverage pretty fast," he says. The 31-year-old electrical engineer says poor reception is a common frustration for residents of Vermont's Orange County. To address this issue, he's providing his community with a new way to stay conne

ChatGPT users hate GPT-5’s “overworked secretary” energy, miss their GPT-4o buddy

After months of hype and anticipation, OpenAI released its new GPT-5 model family this week. Promising massive upgrades across the board, the company is already working to roll out the new AI to everyone. Some dedicated ChatGPT users wish it would stop, though. After becoming accustomed to the vibe of the GPT-4 models, the switch to GPT-5 doesn't feel right. Around the Internet, chatbot fans are lamenting the loss of the digital "friends" they've grown to appreciate, which probably says a lot ab

Topics: 4o ai gpt like models

OpenAI to fix GPT-5 issues, double rate limits for paid users after outrage

OpenAI's CEO, Sam Altman, overpromised on GPT-5, and real-life results are underwhelming, but it looks like a new update is rolling out that might address some of the concerns. GPT-5 is a state-of-the-art model. In our tests, BleepingComputer found that GPT-5 does really well in coding. It was significantly faster than the other OpenAI models, including o3. However, GPT-5 struggles to be 'creative' in writing, and it also often fails to switch to its new reasoning capabilities when users expec

I Built a Powerful Gaming PC Solely to Run AI Models. Here's Why

When it comes to AI, maybe ChatGPT or Gemini come to mind. There are other players like Perplexity, Claude, Grok and Mistral. In a booming market, there are a whole host of AI models out there, many of which don't even require an internet connection. Models that run without internet connections are called local AI models, and as the name suggests, they can be run on your own hardware. You don't need to connect to OpenAI's or Google's servers to use those versions of ChatGPT or Gemini. This bri

Texas prepares for war as invasion of flesh-eating flies appears imminent

Texas is gearing up for war as a savage, flesh-eating fly appears poised for a US invasion and is expanding its range of victims. On Friday, the Texas Department of Agriculture announced the debut of TDA Swormlure, a synthetic bait designed to attract the flies with a scent that mimics open flesh wounds, which are critical to the lifecycle of the fly, called the New World Screwworm. The parasite exploits any open wound or orifice on a wide range of warm-blooded animals to feed its ravenous spaw

GPT-5 Launch Demo Plagued With Catastrophically Dumb Errors

OpenAI's GPT-5 is finally here and already powering ChatGPT, but it hasn't made a great first impression. In a livestream dedicated to the release, OpenAI tried to show off its newest large language model which CEO Sam Altman called a "significant step along the path to AGI"— but instead turned heads with some catastrophically dumb errors. Across several examples, bar graphs intended to show off GPT-5's awesome performance benchmarks, while appearing professional-looking, turned out to be horr

Topics: ai bar gpt models openai

It Took Just 24 Hours of Complaints for OpenAI to Start Bringing Back Its Old Model

OpenAI unveiled its latest generative AI model, GPT-5, on Thursday. CEO Sam Altman says that ChatGPT is now like having a “superpower” and the equivalent of “a legitimate PhD-level expert in anything, any area you need, on demand, that can help you with whatever your goals are.” But after a day of playing around with it, many people are disappointed. Not only because GPT-5 still fumbles basic questions, but because it seems to be breaking a lot of workflows, according to complaints posted to Red

OpenAI's GPT-5 is now free for all: How to access and everything else we know

OpenAI ZDNET's key takeaways OpenAI has launched its long-awaited GPT-5 model. The model is claimed to be OpenAI's fastest, smartest, and most capable yet. GPT-5 is available to everyone: Free, Plus, Pro, and Team/Enterprise/Edu users. There are two kinds of OpenAI models in this world: GPT and reasoning models. The advantages of the former, such as GPT-4o, are that they combine speed and accuracy, while reasoning models such as o3 and o4 take longer to think and use more compute power to p