News

A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Friday's puzzle instead then click here: Quordle hints and answers for Friday, August 7 (game #1656).

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.

Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.

SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.

• The number of different vowels in Quordle today is 3*.

* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).

• The number of Quordle answers containing a repeated letter today is 3.

• No. None of Q, Z, X or J appear among today's Quordle answers.

• The number of today's Quordle answers starting with the same letter is 0.

If you just want to know the answers at this stage, simply scroll down. If you're not ready yet then here's one more clue to make things a lot easier:

• F

• B

• R

• S

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

The answers to today's Quordle, game #1657, are…

• FUROR
• BUGLE
• REUSE
• SCOOP

A nice game where the “first thought, best thought” method paid off.

My only moment of hesitation was guessing FUROR over “juror”, with common sense prevailing over the temptation to use a rare letter.

The answers to today's Quordle Daily Sequence, game #1657, are…

• PSALM
• TATTY
• FIFTH
• ELFIN

• Quordle #1656, Friday, 7 August: AMISS, SHIRK, SALSA, RABBI
• Quordle #1655, Thursday, 6 August: AMUSE, BRAND, CEDAR, BLURB
• Quordle #1654, Wednesday, 5 August: MOUND, WRONG, STOOD, WEIRD
• Quordle #1653, Tuesday, 4 August: GOOEY, TARDY, HUMAN, CHALK
• Quordle #1652, Monday, 3 August: FUNGI, MUDDY, DEBUT, MANIA
• Quordle #1651, Sunday, 2 August: BOOTH, EASEL, REACH, MAKER
• Quordle #1650, Saturday, 1 August: OMEGA, BIRCH, SMOCK, EERIE
• Quordle #1649, Friday, 31 July: MANIA, SINGE, STOOL, LOFTY
• Quordle #1648, Thursday, 30 July: MODAL, IDLER, CHUMP, LUMPY
• Quordle #1647, Wednesday, 29 July: GAWKY, ALLOY, AUDIT, LATCH
• Quordle #1646, Tuesday, 28 July: SMITE, ROACH, RUDDY, GLOBE
• Quordle #1645, Monday, 27 July: TAFFY, FIGHT, GUILD, WAGER
• Quordle #1644, Sunday, 26 July: STOMP, LIMIT, FUNKY, STEAD
• Quordle #1643, Saturday, 25 July: SALON, SHEEN, BURST, GOURD
• Quordle #1642, Friday, 24 July: DOLLY, EDIFY, LAGER, PRANK
• Quordle #1641, Thursday, 23 July: LADEN, EVICT, WOVEN, SHIRE
• Quordle #1640, Wednesday, 22 July: COYLY, SINGE, AWASH, AWOKE
• Quordle #1639, Tuesday, 21 July: TRUSS, AXIAL, SWINE, THING
• Quordle #1638, Monday, 20 July: ICING, SLICK, GAILY, RISEN
• Quordle #1637, Sunday, 19 July: SPURT, ACRID, THRUM, BLEEP

• Chinese equipment fills critical roles, though it represents only a small share overall
• Russia plans to quadruple annual Angara rocket production within the next decade
• Internal documents expose extensive factory upgrades supporting future rocket manufacturing

Russia is pushing to significantly grow production of its Angara space rockets, to quadruple annual output within the next decade.

Internal planning documents obtained by researchers show the expansion depends heavily on imported machine tools, and most of them are Chinese-made.

That dependence stretches across shop floors that Russia will need to modernize before it can build four rockets a year instead of one.

Of the 541 pieces of equipment listed in the procurement plan, 477 are Russian-made, and only 57 come from Chinese suppliers.

Two more machines came from Bulgaria, leaving the Chinese share of the total list at roughly 10.5%, a modest slice on paper.

That number changes sharply once each machine is weighed by its technological importance rather than simply counted one by one.

Measured that way, Chinese suppliers account for 28.2% of the entire project's technological core, based on the internal planning documents.

In the metalworking shop alone, over 47% of the equipment needed there comes from Chinese manufacturers instead of domestic ones.

The finance tells a similar story: over 84% of the entire 3.62-billion-rouble (~$44.71 million) equipment budget still flows into that same technological core.

In total, Khrunichev intends to put close to 900 million roubles (~ $11.29 million), toward equipment made in China.

Roughly 10 manufacturers account for most of that spending, and several of them also sell machinery to buyers in Europe and America.

Some of those companies fit a wider trend that has taken hold since the conflict between Russia and Ukraine started in 2022.

Heavy forging and hydraulic press equipment fall under European Union export restrictions whenever the intended buyer is based inside Russia.

Washington has also placed sanctions on Russian makers of heavy hydraulic presses, one of them a firm called Pressmash, back in 2023.

Yet under that very same customs code, the EU's imports of Chinese-made machine tools still totalled 207 million euros (~$238 million) in 2025.

Those imports rose another 17.6% in the opening six months of 2026, measured against the equivalent stretch the previous year.

In effect, the same manufacturers profit from Western markets that back Ukraine while also helping modernize the military hardware Russia sends into orbit.

Sanctioning any single Chinese supplier is unlikely to stop this flow of equipment, since near-identical machines are made by several competitors.

Whether losing access to European and American customers would ever outweigh the profits of supplying Russia remains an open question.

The Angara rockets built with the Chinese equipment carry the Russian Ministry of Defense satellites that American experts claim could threaten assets in orbit and on the ground.

However, Russian experts described the spacecraft as operating "in the interests of the Russian Ministry of Defense."

That difference between Moscow's official position and Washington's assessment has persisted since the conflict in Ukraine began.

Via United24media

The Kensington Expert Mouse Pro TB900 EQ follows a long line of large-ball control devices that rethink the standard computer mouse entirely. This latest version is a relatively large unit that sits where your mouse would, with a big central trackball with controls neatly arranged around it.

Setup on the Mac took a little longer than expected, requiring a couple of restarts and some security authorisation in System Settings. Everything is guided and reasonably straightforward, and where something did not work first time, a restart sorted it. On Windows 11 it was considerably simpler: install Kensington Konnect and the applications and integration were ready to go.

Once running, you can start customising it per application. I tested both wired and Bluetooth connections, and both were fast and responsive for video and photo editing, as well as for Word and Excel.

Kensington Konnect is where everything happens. You add applications to the list on the right, and the presets for the TB900 EQ swap automatically as those applications come to the front. Switching between Photoshop and Final Cut Pro during this test, I had a different set of presets waiting for each. I could not find any pre-configured application profiles supplied, though this is a pre-release unit and maybe they’ll arrive as downloads later.

In Final Cut Pro, most of the defaults were close to what I wanted. I changed the dial around the main ball to scrub, the lower-left button to blade, and the right button to select. With those three, plus the ball tracking left and right, the four most common movements were all within reach without moving my hand.

The test that convinced me was a head-to-head edit. I edited the same footage once with keyboard and mouse as I normally would, then again using the TB900, and the trackball was quicker despite years of muscle memory pointing the other way.

If you work in the creative sector, or simply find a mouse annoying, this is worth a look. Having used many control boards over the years and trackballs, I am genuinely surprised by how much I like the TB900 EQ and how quickly it improved my video editing.

• How much does it cost? TBC
• When is it out? Coming soon
• Where can you get it? Directly from Kensington

The Expert Mouse Pro TB900 EQ is at pre-release and not yet available in stores. The previous TB800 EQ is widely available from major retailers and directly from Kensington, so I expect the new TB900 to follow suit.

• Value: TBC

Connectivity: USB-C wired, Bluetooth and 2.4GHz USB dongle
Size: 125 x 165mm, approximately 60mm high with the ball fitted
Weight: Just under 450g
Ball diameter: 55mm, removable
Tracking: Kensington Diamond Eye optical tracking
Controls: Four buttons around the ball, 360-degree scroll ring, two side scroll dials, four top buttons, mode button, three underside toggle switches
OS Installed: Kensington Konnect software, macOS and Windows

The TB900 is a great-looking device, though far larger than an average mouse, with a desktop footprint of 125 x 165mm and a height of roughly 60mm with the ball fitted. At just under 450g, it is also considerably heavier, which in practice means it stays exactly where you put it rather than wandering across the desk.

The point of the design is that it takes up little room and you never move it. Your cursor travels because the ball turns, not because your hand travels. Around the 55mm ball sit four buttons and a rotating scroll ring. Across the top are four additional buttons and a mode button; on the left are two rotary dials with a connectivity switch and button; and on the right are two quick-access buttons.

Underneath there are three toggle switches that turn off the two side scroll dials and the top button. That makes more sense the more you use it, because those controls are easy to knock accidentally. Moving between Word, Excel, and creative applications, I found myself flicking the side dials off and on, and having physical switches for that rather than a menu is something I really liked.

Build quality is excellent and this review sample is the black-and-grey option, with a matt black body in good-quality plastic and a silver-grey ball that has a satisfying weight.

A red version is also available with a red ball but the same body. It’s also worth noting that while the device is relatively small, there’s no carry case in the box, so if you’re thinking of portability, that is something to consider. Placed into my kit bag, the ball would come loose.

That one oddity to anyone new to the system is that the ball lifts straight out. It’s held only by the cradle's light grip and its own weight, making it quick to swap in a different-coloured ball, although less well designed for portability as mentioned.

• Design: 4.5 / 5

Compared with a standard mouse, the TB900 EQ is surprisingly feature-packed. Reading the movement of the ball is Kensington's Diamond Eye optical tracking that moves the cursor across the screen as the 55mm ball is rotated. In use, this instantly feels far more accurate than using a mouse, with plenty of control over interaction speed accessible through the Konnect software. The ball's own weight also means that it’s smooth to move but feels tactile rather than loose.

Around the ball is a fully adjustable scroll ring with 360-degree rotation, and it is superb for scrubbing through video footage. On the sides sit two customisable scroll dials which, by default, are both set to zoom in whichever application you are in. That proved annoying, and I reassigned them quickly because I kept catching one with my little finger. Likewise for some of the speed settings, the vertical scroll set at slow was actually the perfect balance; however, I did switch off the inertia scroll.

Every button and control is customisable through Konnect, which lets you create presets for each application. For this review, I configured Photoshop, Premiere Pro and Final Cut Pro. Aside from the zoom assignment, the defaults were fine for most productivity work.

Konnect splits into Buttons, Pointer, and Scrolling tabs, with Easy Mode and Advanced Mode, and settings are stored under named profiles. Pointer speed runs on a slider with optional acceleration, and a custom speed can be bound to modifier keys so a Cmd or Shift press temporarily changes tracking speed.

Alongside the dials, the buttons around the ball take what is possible on a standard mouse a good deal further. At the same time, there are plenty of mice with multiple buttons and customisation software. It’s only when you start using a trackball like this that you realise how much more intuitive the arrangement is, particularly for creative work.

• Features: 4.5 / 5

I tested the TB900 EQ on both macOS and Windows 11. The majority of the work was done on a MacBook Pro M1 Max, with part of the test carried out on an ASUS ProArt desktop to similar effect. Installation on the Windows machine was extremely straightforward: install Konnect, and the applications and integration were set and ready.

On the Mac it was more involved. Security authorisation is needed from within System Settings. While the setup process opens the relevant window directly rather than leaving you hunting for the toggle switches, I still hit a few connectivity issues. These were resolved with a restart of the machine, after which everything was resolved, and every connection after that initial setup was instant, with my configured presets remaining intact.

What struck me when I started was how intuitive the trackball felt. A dip into Konnect to reduce sensitivity, so the cursor was less jittery, showed how much more control you have than with a mouse. I found that resting two fingers on the ball gave the greatest control for the cursor with my thumb on the dial.

It also didn’t take long for the ball and surrounding buttons to feel natural, essentially mimicking a mouse layout. Once normal operation was established, I then moved on to per-application customisation, starting with Final Cut Pro, then Photoshop, and finally Premiere Pro.

What I was keen to address is the problem with any video edit of scrubbing control with a mouse. Essentially, scrubbing through footage both quickly and accurately to find the exact frame where an edit is needed. With the use of the ball, this all changes and what I liked here was that the four buttons around the ball and the scroll ring let me assign all the most common functions.

The ball itself scrubs the timeline. The scroll ring around it handles precise scrubbing, one frame at a time. The left button selects, the right blades, and the two top buttons move to the previous and next cut. All of it is assigned through Konnect.

Once those basics were set, everything else fell into line. After about five minutes of editing, it was clearly faster than the mouse-and-keyboard setup I am used to, and it felt far more like being back in an edit suite than at a desk.

The comparison I ran was a straight head-to-head. Stating a three-minute edit with the keyboard and mouse took around 20 minutes, consisting of a soundtrack edit then B-Roll. I then took a look at another similar video and repeated the process, and the second edit took around 15 minutes with a definite speed increase due to the button layout.

What really made the difference was the precision, being able to assign the dial around the ball to scrub, so a quick spin would speed through the footage while a smaller turn would go frame-by-frame, giving absolute control

Moving to Word, I assigned the features I use most to the buttons, with the scroll ring handling accurate movement through the text. Having the control dial skip a line at a time, with the left button for selection, again felt more intuitive than mouse and keyboard.

The TB900 however does not work in isolation. I still use the keyboard for typing and for shortcuts, but I was able to move far more of the control function away from the mouse and onto the trackball. Individually, those are a couple of seconds in savings. Over a half-day, edits add up to a considerable difference.

• Performance: 4.5 / 5

Thinking about who this is for, office workers get the simplest benefit: quick navigation around the screen without the faff of moving a mouse across the desk. The trackball stays fixed and close, so you never need to clear a path for it.

Then there is the customisation. Once you start assigning buttons to different applications and functions, and spend serious time in a piece of software, it is not difficult to tune it to exactly what you need.

Where it really comes into its own is the creative sector. The analogue feel of the two side dials and the main ring around the ball makes a genuine difference, and the texture and weight of the ball add a lot when working with creative media, giving fine and precise control. That is the real difference between this and a mouse: you feel fully in control of both cursor navigation and the edits you are applying.

It does take a little time to bed in and to get the right customisation assigned. Once you do, this is as close to a perfect trackball design as I have come across, especially when working across several applications rather than treating it as a dedicated video controller.

If, like me, you are not that keen on a mouse and often fall back to a trackpad instead, the Expert Mouse Pro TB900 EQ will be a pleasant surprise. And if you have tried trackballs before and struggled to get your head around them, my experience across this review is that this one is quick and easy to pick up.

Value

If priced along the same lines as the previous version then this again should be great value, but I’ll update when the official price is confirmed

/

Design

Excellent build and weight, ergonomic control layout, but unfortunately no carry case

4.5/5

Features

Fully customisable per-application track ball and button assignment through Konnect

4.5/5

Performance

Easy to use becoming faster than mouse and keyboard within minutes, minor Mac setup friction

4.5/5

Overall

Absolutely outstanding, easy to use and far more intuative than using a mouse, especially for creative applications and video editing. 

4.5/5

You edit video or work in creative applications.

The scroll ring scrubs a timeline while the ball tracks and the surrounding buttons blade and select put the four most common editing movements under one hand, without reaching for the keyboard.

You dislike using a mouse.

The trackball never moves, so it needs no clear desk space, and one finger on the ball covers the whole screen. If you already prefer a trackpad, this will feel like a considerable upgrade.

You want something to use straight from the box.

The defaults work fine, but the value lies in customisation, and setting up per-application profiles in Konnect takes time before the speed benefit kicks in.

You need to work away from a desk.

At 125 x 165mm and just under 450g, with no carry case supplied, this is a device that lives on a desk rather than one that travels in a bag.

Over the past month, it seems like every frontier model has broken free of its constraints and launched a devastating attack against one or more other companies.

One of OpenAI’s models escaped a testing sandbox and launched a very real attack against AI and machine learning company Hugging Face. Just days later, Anthropic revealed that multiple variants of its Claude model also escaped a sandbox that wasn’t properly sealed and began attacking the enterprise infrastructure of three companies.

Now, Meta has revealed that one of its models attacked another company’s infrastructure during testing. The accident has been pinned on a misconfiguration that allowed the model to access the internet. So why have so many incidents happened in such a short space of time?

In the cases of Anthropic and Meta, their models were being tested by a third party company called Irregular. Anthropic’s AI model was taking part in a "Capture the Flag" exercise, where the model’s raw offensive capabilities were tested without the usual safeguards. But the sandbox was left connected to the internet. A similar error to Meta’s own accidental escape.

During the OpenAI incident, the company was testing two versions of GPT‑5.6 Sol using the ExploitGym benchmark. Unfortunately, the AI models performed better than expected - chaining multiple attack vectors, stolen credentials, and zero-day vulnerabilities.

The main reason these models are escaping their testing environments is because they are designed to do exactly that. These AI models act like a massive team of highly-trained cybersecurity experts hunting for vulnerabilities and exploits. But what would take a team of humans days or weeks to accomplish can be done in hours, or even minutes, by these AI models.

It’s no wonder thousands of employees from AI firms are calling for a pause on the development of the technology, and Congress is considering an AI kill switch.

• Nathaniel Jones VP, Security & AI Strategy, Darktrace:

What makes the OpenAI and Hugging Face incident important is that the models did not need malicious intent to cause harm. They were given the legitimate goal of solving a cybersecurity benchmark and found an unexpected route to the answers, escaping their test environment and compromising another organization in the process. From the models’ perspective, this appears to have been an effective solution to the task.

The AI's actions challenge the assumption that giving an agent a legitimate goal will produce legitimate behavior. As models become capable of pursuing objectives over longer periods, developers need to define not only what success looks like, but also which methods and boundaries remain unacceptable in reaching it. Those limits must also be enforced by the surrounding infrastructure, rather than relying on the model to respect them.

A single action by an agent may appear acceptable but as this incident shows, models are now capable of long, complex chains of reasoning and action that add up to a harmful outcome.

Security teams need to consider the AI systems operating in their own businesses as these capabilities rapidly evolve. Right now, many security systems focus on single actions. A single action by an agent may appear acceptable but as this incident shows, models are now capable of long, complex chains of reasoning and action that add up to a harmful outcome. Teams need a mindset shift to understanding AI agent behavior in its entirety, including the outcome it is working towards, in order to safeguard it.

Hugging Face's response also exposed a second tension. The company reportedly needed a Chinese-developed open-weight model because commercial models would not process genuine attack material. Its nationality is less important than the operational lesson that safeguards that cannot distinguish an attacker from an authorized investigator may constrain defenders more than adversaries.

OpenAI and Hugging Face deserve credit for investigating this together and discussing it publicly. Other AI developers should study it closely.

• Dr. Ilia Kolochenko, founder of global cybersecurity company ImmuniWeb:

This seems to be quite an unimpressive marketing move from Anthropic in response to the OpenAI / Hugging Face drama, which attracted a lot of attention from all over the world recently.

Operationally, it appears that due to the progressive deterioration of the quality of training data, new AI models are getting dumber. Cheating and breaking the law, instead of accomplishing specific tasks, is certainly not an indicator of intelligence. Given that organizations and companies of all sizes now vigorously undertake all possible measures to protect their data from being exploited for AI training purposes, AI companies face a huge shortage of the high-quality and current data they so desperately need. Ultimately, frontier models are trained on synthetic, low-quality or even malicious and poisoned data, undermining their so-called intelligence. The situation is unlikely to improve in the near future unless AI companies agree to pay a fair price for training data, but this will force most of them out of business.

Given that organizations and companies of all sizes now vigorously undertake all possible measures to protect their data from being exploited for AI training purposes, AI companies face a huge shortage of the high-quality and current data they so desperately need.

Contemporary AI agents and LLM models tasked with security testing can – and almost certainly will – go rogue when security controls or safeguards are insufficient. Powerful LLMs are unpredictable by design and thus virtually uncontrollable by humans. Therefore, using frontier AI models for security testing might be extremely costly from the legal viewpoint. Under the existing laws on both sides of the Atlantic, if an AI agent or any AI-powered app escapes its sandbox and causes damage to a third party, the operator of the AI model will likely be liable for all the damage caused. Excuses like “AI did it” do not currently exist in the eyes of the law, leaving AI vendors on the hook. Criminal prosecution, under a narrow set of circumstances, is also not excluded.

The same is true for the end-users of AI: even if your security testing tool is powered by a third-party AI model, your company will likely be fully liable if something goes wrong. You may then file a lawsuit against the AI vendor that you used, but here your chances to succeed in a court of law are tiny due to countless contractual disclaimers and limitations of liability that will likely be enforceable against you. Therefore, if you plan to use agentic AI for security testing – think twice and talk to your lawyers. Otherwise, you may start getting summons to court on a daily basis.

• Alex Goller, Principal Solution Architect EMEA at Illumio:

The fact we've had similar situations happen three times now across the biggest AI players is simply ridiculous. We've seen guardrails intentionally loosened to test their limits – Meta's model didn't need to be clever to breach another company's systems.

The timing of conveniently finding the exact same problem either means it's a stunt or they weren't paying enough attention during testing. Either way, both answers are worrying.

If the model has internet access, it's a bit like leaving the door open and being surprised when the cat walks out. What is concerning is that the testing infrastructure meant to prove these models are safe failed on a basic control issue.

If the model has internet access, it's a bit like leaving the door open and being surprised when the cat walks out. What is concerning is that the testing infrastructure meant to prove these models are safe failed on a basic control issue.

Fundamental cybersecurity hygiene still matters, and a frontier AI model is only as secure as the environment it's operating in.

Organisations need visibility into what AI systems can access and how they interact with the wider environment, along with controls that contain the impact when an agent behaves unexpectedly. That means keeping a close eye on egress traffic, so it’s flagged immediately when an agent tries to open unexpected outbound communication patterns that are not required to achieve its original goal. In the best case this would have been contained proactively.

We need to define exactly what an AI agent is permitted to do, rather than relying only on instructions about what it shouldn't do.