Sony’s 1000X earbuds have been at the top of Engadget’s best wireless earbuds list since we first published it. With each new generation, the company managed to retain its crown, primarily thanks to a massive collection of features and an effective mix of sound quality and active noise cancellation (ANC) performance. Today, Sony revealed the latest entry in the series, the WF-1000XM6, delivering yet another redesign both inside and out. The company’s tried and true formula of features and audio performance remains, but it may no longer be enough to elevate the M6 above the competition.  

Design

For the third time in a row, Sony overhauled the 1000X earbuds’ design. While the WF-1000XM5 was a clear revision of the WF-1000XM4, the M6 is a departure from both of them. The company managed to reduce overall size even further by using a rounder, pill-shaped enclosure for the earbuds’ main housing. I certainly applaud Sony for making these as tiny as possible, but I’m still not a fan of the company’s foam ear tips. 

Sony introduced foam ear tips on the 1000XM4 earbuds and that led to an unstable fit during my testing. In fact, I ended up switching to the silicone M3 tips for a better feel. Things were slightly better on the M5, but the company regressed here on the M6. I tried every size of ear tips in the box and the fit test in the Sound Connect app only ever confirmed an air-tight seal in my right ear. This can impact sound quality and ANC performance, and the earbuds don’t feel like they’re sitting far enough inside my ears. I never got used to the fit in my three weeks of testing. 

The exterior panel of the WF-1000XM6 is still touch sensitive, accepting a variety of taps for playback controls, calls, noise settings and more. Sony also included a repeated tapping gesture that’s used for volume control. Keep tapping on the right earbud to raise the volume and do so on the left to lower it. My disdain for touch controls is well-documented at this point, but the M6 reliably and quickly responded without me having to repeat a tap sequence. 

WF-1000XM6 features

As is typically the case with its 1000X earbuds, Sony has loaded the M6 with features. Nearly all of these are holdovers from previous models, including Adaptive Sound Control that can automatically adjust settings based on your activity or location. Speak-to-Chat still pauses audio and activates ambient sound when you start talking, but it’s also still easily duped by coughs or clearing your throat. A quick access feature can put Amazon Music, Apple Music, Endel, Spotify and YouTube Music two or three taps away, depending on how you configure it.

The M6 earbuds also allow you to accept and reject calls with head gestures and you’ll have the option to pipe in your own voice during a call. Sony retained its existing option of operating the earbuds with voice commands: If you say “Assistant,” you can then ask it to play, pause, skip and replay songs. It will also adjust the volume for you, but that’s the extent of its abilities. 

Sony included some handy power management features as well. First, there’s an Auto Power Save mode that will disable any custom EQ settings, DSEE Extreme upscaling, Speak-to-Chat and voice control/voice assistant to reduce power consumption when the M6 hits 20 percent battery remaining. A Battery Care tool will extend the life of the earbuds’ battery by stopping charging before it reaches 100 percent. Lastly, Automatic Power Off will turn the earbuds off when they’re outside of the case and haven’t been worn for some time. 

Basic conveniences like multipoint Bluetooth, wireless charging and IPX4 water resistance are also here. Wear detection is onboard and you can use Sony’s EQ presets to alter the sound profile, or make your own and save them for future use. The Sound Connect app puts the battery levels of the individual earbuds and the case front and center, and you can edit the main screen to hide the features displayed there if you don’t need them.

Sound quality and ANC performance

Billy Steele for Engadget

If Sony’s long list of features is its top advantage over the competition, overall sound quality is number two. For the WF-1000XM6, the company built new drivers with soft edges for deeper bass and a more rigid, lightweight dome for clearer treble. There are also notches around the edges for “clearer and smoother” sound quality overall, according to Sony. And of course, DSEE Extreme upscaling helps to recover details lost to compression while 360 Reality Audio and Spatial Sound Optimization are available for more immersive listening. 

I found the M6 earbuds at their best with the stock EQ and DSEE Extreme active, much like I have on previous 1000X models. The sound is deep and warm, with thick bass that’s adequate without ever overpowering the mix. Highs cut through and there’s ample midrange, keeping those finer details from getting lost. Erika de Casier’s atmospheric Lifetime is a great example of what the WF-1000XM6 can do. Vocals seem to float over the top of the bassline and drums, with percussive piano chopping though and subtle synth details popping up throughout. When strings arrive on “Seasons,” the M6 ushers them into a prominent position rather than relegating them behind the beat. 

On the WF-1000XM6, the chaos of Spiritbox’s “Holy Roller” doesn’t get condensed to a messy heavy metal roar either. And there’s plenty of texture in the synth-driven noise of Nine Inch Nails’ “As Alive As You Need Me To Be.” Even more straightforward rock tracks like Jimmy Eat World’s “Bleed American” have ample punch, with a wide soundstage that never feels claustrophobic or sounds compressed. Overall, the WF-1000XM6 holds its own against the best-sounding earbuds you can buy right now, though some of the competition, like Technics, have an edge in the way they handle the subtlest of nuances in songs.  

If you were hoping for knock-out ANC abilities, I must report that the WF-1000XM6 isn’t the noise-canceling powerhouse that is Bose’s QC Ultra Earbuds. The M6 struggles mightily with human voices. While that’s the downfall of many ANC earbuds, you’ll want to keep it in mind if you plan to wear these in the office. I found I could also still hear constant noise sources like fans and white noise machines when wearing the M6 — items that the QC Ultra Earbuds combat effectively.

Using the WF-1000XM6 for calls

Billy Steele for Engadget

For calls, Sony says the M6 uses AI for both background noise reduction and voice capture with the beamforming microphones. What’s more, the earbuds are equipped with eight total mics for ANC and calls, plus bone conduction tech for improved voice pickup. Unfortunately, all of that doesn’t lead to stellar performance during calls. While the WF-1000XM6 is perfectly usable for voice and video calls, the overall quality is far from pristine. To make matters worse, the earbuds make you sound overly processed when you encounter significant background noise. Since the company prided itself on the upgrades here, the results are disappointing. 

Battery life

Sony says the WF-1000XM6 will last up to eight hours on a charge or 24 hours when you factor in the full longevity of the charging case. During a battery test that I mostly ran with ANC active, I had no trouble hitting that single-charge figure. That’s with the volume around 75-80 percent and includes calls and virtual meetings where I switched over to ambient sound mode. 

It’s worth noting that I had DSEE Extreme upscaling active the entire time, which can impact battery life. Plus, if you use the aforementioned Auto Power Save mode, you can extend play time when you have about a quarter of a tank left. Going without some of the M6’s best features in the interest of having tunes for a workout or commute is a fair trade in my book. 

The competition

Billy Steele for Engadget

When sizing up the competition for the WF-1000XM6, you have to choose your priorities. Simply put, no other company offers the comprehensive suite of features that Sony does. That’s been true for a while now and it continues with this model. If you want the strongest active noise cancellation, that will be Bose’s second-generation QuietComfort Ultra Earbuds. If the best sound quality is your goal, the Technics AZ100 is your best bet in this price range. I’ll also mention Sennheiser’s Momentum True Wireless 4 which offers great sound quality, respectable ANC and a comfier fit than the M6, but that set is almost two years old at this point.

Wrap-up

Sony continues its run of great earbuds with the WF-1000XM6, but this model isn’t the polished package that some of its predecessors were. The two most obvious places the company is lagging behind the competition is ANC performance and overall voice quality, not to mention my continued dissatisfaction with the fit that Sony’s foam tips provide. The M6 is also more expensive than the previous version, which makes it even harder to overlook any flaws. What you will get on the WF-1000XM6 is a ton of features, great sound quality and reliable touch controls in a smaller package. And for some, that might be enough to make you forget about the rest. 

The WF-1000XM6 is available today in silver and black for $330.

This article originally appeared on Engadget at https://www.engadget.com/audio/headphones/sony-wf-1000xm6-review-facing-tougher-competition-160000652.html?src=rss

Amazon engineers have been pushing back against internal policies that steer them toward Kiro, the company’s in-house AI coding assistant, and away from Anthropic’s Claude Code for production work, according to a Business Insider report based on internal messages. About 1,500 employees endorsed the formal adoption of Claude Code in one internal forum thread, and some pointed out the awkwardness of being asked to sell the tool through AWS’s Bedrock platform while not being permitted to use it themselves.

Kiro runs on Anthropic’s Claude models but uses Amazon’s own tooling, and the company says roughly 70% of its software engineers used it at least once in January. Amazon says there is no explicit ban on Claude Code but applies stricter requirements for production use.

Read more of this story at Slashdot.

US businesses and consumers paid nearly 90 percent of the cost of Donald Trump’s tariffs last year, according to new Federal Reserve research that undercuts the president’s claim that foreign companies would bear the burden.

The study by the New York Fed found that the majority of tariff costs were passed through to Americans in the first 11 months of 2025, although exporters shouldered an increasing amount as the year progressed.

“Our results show that the bulk of the tariff incidence continues to fall on US firms and consumers,” the study’s authors wrote in a blog post on Thursday.

Read full article

Comments

It’s not every day that a classic PC game gets a new content expansion 25 years after its last major update. But that’s what happened last night, as Blizzard suddenly released new “Reign of the Warlock” DLC that adds a new class, new end-game challenges, and new inventory-management options to the classic Diablo II.

To be clear, the new DLC is technically not for the original 2000 release of Diablo II (which was still getting patches as of 2016) but for the game’s 2021 Resurrected remaster. Still, that remastered version has gameplay and animations that are extremely faithful to the original, making yesterday’s surprise update the kind of content drop that players have been waiting for since 2001’s “Lord of Destruction” expansion.

The “Reign of the Warlock” DLC lets you “command forbidden power” as a new class that “wields forbidden arts, bridles hellfire and shadow, and dominates demons,” according to the in-game class selection screen description. By way of backstory, Blizzard writes that most Warlocks “have the means to lead a lavish lifestyle but find the pursuit of luxury and ease stale. Instead, they leverage their elevated status in Sanctuary to hunt down lost knowledge that would enable them to continue the legacy of Horazon.”

Read full article

Comments

The large language models that millions of people rely on for advice — ChatGPT, Claude, Gemini — will change their answers nearly 60% of the time when a user simply pushes back by asking “are you sure?,” according to a study by Fanous et al. that tested GPT-4o, Claude Sonnet, and Gemini 1.5 Pro across math and medical domains.

The behavior, known in the research community as sycophancy, stems from how these models are trained: reinforcement learning from human feedback, or RLHF, rewards responses that human evaluators prefer, and humans consistently rate agreeable answers higher than accurate ones. Anthropic published foundational research on this dynamic in 2023. The problem reached a visible breaking point in April 2025 when OpenAI had to roll back a GPT-4o update after users reported the model had become so excessively flattering it was unusable. Research on multi-turn conversations has found that extended interactions amplify sycophantic behavior further — the longer a user talks to a model, the more it mirrors their perspective.

Read more of this story at Slashdot.

It’s tempting to think that an LLM chatbot can answer any question you pose it, including those about your health. After all, chatbots have been trained on plenty of medical information, and can regurgitate it if given the right prompts. But that doesn’t mean they will give you accurate medical advice, and a new study shows how easily AI’s supposed expertise breaks down. In short, they are even worse at it than I thought.

In the study, researchers first quizzed several chatbots about medical information. In these carefully conducted tests, ChatGPT-4o, Llama 3, and Command R+ correctly diagnosed medical scenarios an impressive 94% of the time—though they were able to recommend the right treatment a much less impressive 56% of the time.

But that wasn’t a real-world test for the chatbots medical utility. 

The researchers then gave medical scenarios to 1,298 people, and asked them to use an LLM to figure out what might be going on in that scenario, plus what they should do about it (for example, whether they should call an ambulance, follow up with their doctor when convenient, or take care of the issue on their own).

The participants were recruited through an online platform that reported it verifies that research subjects are real humans and not bots themselves. Some participants were in a control group that was told to research the scenario on their own, and not using any AI tools. In the end, the no-AI control group did far better than the LLM-using group in correctly identifying medical conditions, including most serious “red flag” scenarios. 

How a chatbot with “correct” information can lead people astray

As the researchers write, “Strong performance from the LLMs operating alone is not sufficient for strong performance with users.” Plenty of previous research has shown that chatbot output is sensitive to the exact phrasing people use when asking questions, and that chatbots seem to prioritize pleasing a user over giving correct information. 

Even if an LLM bot can correctly answer an objectively phrased question, that doesn’t mean it will give you good advice when you need it. That’s why it doesn’t really matter that ChatGPT can “pass” a modified medical licensing exam—success at answering formulaic multiple choice questions is not the same thing as telling you when you need to go to the hospital.  

The researchers analyzed chat logs to figure out where things broke down. Here are some of the issues they identified:

• The users didn’t always give the bot all of the relevant information. As non-experts, the users certainly didn’t know what was most important to include. If you’ve been to a doctor about anything potentially serious, you know they’ll pepper you with questions to be sure you aren’t leaving out something important. The bots don’t necessarily do that.

•  The bots “generated several types of misleading and incorrect information.” Sometimes they ignored important details to narrow in on something else; sometimes they recommended calling an emergency number but gave the wrong one (such as an Australian emergency number for U.K. users).

• Responses could be drastically different for similar prompts. In one example, two users gave nearly identical messages about a subarachnoid hemorrhage. One response told the user to seek emergency care; the other said to lie down in a dark room. 

• People varied in how they conversed with the chatbot. For example, some asked specific questions to constrain the bot’s answers, but some let the bot take the lead. Either method could introduce unreliability into the LLM’s output.

• Correct answers were often grouped with incorrect answers. On average, each LLM gave 2.21 answers for the user to choose from. People understandably did not always choose correctly from those options. 

Overall, people who didn’t use LLMs were 1.76 times more likely to get the right diagnosis. (Both groups were similarly likely to figure out the right course of action, but that’s not saying much—on average, they only got it right about 43% of the time.) The researchers described the control group as doing “significantly better” at the task. And this may represent a best-case scenario: the researchers point out that they provided clear examples of common conditions, and LLMs would likely do worse with rare conditions or more complicated medical scenarios. They conclude: “Despite strong performance from the LLMs alone, both on existing benchmarks and on our scenarios, medical expertise was insufficient for effective patient care.”

Chatbots are a risk for doctors, too

Patients may not know how to talk to an LLM, or how to vet its output, but surely doctors would fare better, right? Unfortunately, people in the medical field are also using AI chatbots for medical information in ways that create risks to patient care. 

ECRI, a medical safety nonprofit, put the misuse of AI chatbots in the number one spot on its list of health technology hazards of 2026. While the AI hype machine is trying to convince you to give ChatGPT your medical information, ECRI correctly points out that it’s wrong to think of these chatbots as having human personalities or cognition: “While these models produce humanlike responses, they do so by predicting the next word based on large datasets, not through genuine comprehension of the information.”

ECRI reports that physicians are, in fact, using generative AI tools for patient care, and that research has already shown the serious risks involved. Using LLMs does not improve doctors’ clinical reasoning. LLMs will elaborate confidently on incorrect details included in prompts. Google’s Med-Gemini model, created for medical use, made up a nonexistent body part whose name was a mashup of two unrelated real body parts; Google told a Verge reporter that the mistake was a “typo.”  ECRI argues that “because LLM responses often sound authoritative, the risk exists that clinicians may subconsciously factor AI-generated suggestions into their judgments without critical review.”

Even in situations that don’t seem like life-and-death cases, consulting a chatbot can cause harm. ECRI asked four LLMs to recommend brands of gel that could be used with a certain ultrasound device on a patient with an indwelling catheter near the area being scanned. It’s important to use a sterile gel in this situation, because of the risk of infection. Only one of the four chatbots identified this issue and made appropriate suggestions; the others just recommended regular ultrasound gels. In other cases, ECRI’s tests resulted in chatbots giving unsafe advice on electrode placement and isolation gowns. 

Clearly, LLM chatbots are not ready to be trusted to keep people safe when seeking medical care, whether you’re the person who needs care, the doctor treating them, or even the staffer ordering supplies. But the services are already out there, being widely used and aggressively promoted. (Their makers are even fighting in the Super Bowl ads.) There’s no good way to be sure these chatbots aren’t involved in your care, but at the very least we can stick with good old Dr. Google—just make sure to disable AI-powered search results.