from the acceptable-collateral-damage,-I-assume dept

ICE never needed officers to disguise themselves with masks and strip themselves of identification before Trump took office for the second time. What ICE is doing now isn’t what ICE was doing during Trump’s first term, even though it’s the same hateful bigot sitting behind the Resolute Desk he thinks should be covered in gold leaf.

According to the DHS, ICE officers need to look like roving kidnapping squads because they fear for their safety. Supposedly, they’re under attack now more than ever, something not even supported by the DHS’s context-free claims of massive increases in assaults of ICE officers.

ICE has never been popular. People have been calling for ICE to be abolished for far longer than the last 18 months of its existence. But now that ICE behaves like an invading force, rather than an agency involved in immigration and customs enforcement, more people are reacting to its unwanted presence in their neighborhoods.

ICE’s excuses for mask-wearing were [cough] unmasked when ICE was asked to fill in for unpaid TSA agents. ICE officers showed up at airports without masks to stand around and milk the clock, apparently unworried about being “exposed” or subjected to threats to them or their families.

But now that the TSA is as staffed as it’s ever going to be, ICE is returning to American streets, long on masks and short on training. Criminal opportunists know a good thing when they see it. When it’s impossible to tell whether the person assaulting you/demanding access to your home/running off with your valuables is an actual federal officer or just someone with access to ski masks and camo, the criminals have the upper hand.

As of February, Noticias Telemundo had documented at least six cases of impostors posing as ICE agents to rob or harass immigrants. In mid-January, a man broke into a house in Pittsburgh claiming to be an ICE agent and threatening a teen with a knife. In February, police in San Diego said a man allegedly impersonated an officer and wrapped his arms around the neck of a restaurant manager, claiming the manager was in the country illegally and he was going to arrest him.

Sure, some of you may be scoffing at “six cases” since Trump won the election. But that’s only the ones where a (foreign!) news agency managed to put together the pieces to deliver reporting that should have been done much earlier by domestic new agencies.

Here’s the more damning stat:

Of the 31 impersonation cases documented in 2025, 84% involved individuals who claimed to be ICE agents. Others identified themselves as officers from Border Patrol or the Department of Homeland Security.

Thirty-one impersonations. Apparently all of them involved people pretending to be in the business of migrant deportation. And it’s not just the normal crime you’d expect from criminals seeing a flaw in the system and exploiting it. It’s also led to an increase in the sort of crime this administration will likely greet with pardons and payout from the “FUCK AMERICA $1,776 MILLION SLUSH FUND.”

The recorded incidents include intimidation, robbery and sexual assault, as well as so-called “immigration operations” carried out by armed vigilantes against what they describe as an “invasion” of foreigners in the U.S.

This was a problem the FBI recognized months ago, but rarely speaks of now because it’s being led by the only guy who has a chance at drinking Defense Department Secretary Pete Hegseth under the table. The current “leadership” has nothing to say about giving criminals more opportunities to engage in criminal acts.

Neither DHS nor ICE responded to Noticias Telemundo’s request for official statistics about cases of fake ICE agents. They also did not comment on the trends revealed by this investigation.

Not even the rote “fake news” quasi-rebuttal from this miserable assortment of inhuman asshats. Well, if DHS and ICE won’t speak for themselves, I’ll let this next quote from NBC/Telemundo speak for itself:

“You’re going back to Mexico,” a man told the immigrants in a video recorded from inside their truck. He insulted them for their appearance and for not speaking English, took their keys and snatched the immigrant’s phone when he called his boss. The manager later told the police that the fake agent had claimed to be from ICE and had warned him that all his employees were going to go to “f—–g jail.”

This isn’t fake news. This isn’t implication extrapolated from minimal inference. There are literal recordings of these impersonations.

This isn’t people imagining the worst because they’re politically opposed to the current administration. These are documented instances of the only thing that could be worse than the brutality and bigotry perpetrated by this administration: criminal acts encouraged by this government’s unwillingness to do its dirty work honestly.

Filed Under: bigotry, dhs, ice, masked officers, mass deportation, thugs, trump administration

Honorable Mentions

Photograph: G Stockstudio/Getty Images

As we said, WIRED runners pound hundreds of miles every year. Here are a few of the other shoes we’ve tested that you might want to consider if the above do not work for your foot. If you’re not familiar with a brand, we recommend going to a local running store for a test run before plunking down your credit card.

Diadora Nucleo 2 for $165: The Nucleo 2 isn’t a wow, high-energy, super springy shoe. But if you’re a fan of straightforward, no nonsense comfort and good inherent stability across a good range of paces, the Nucleo 2 delivers.

Rad R1 for $130: Made to master gym, HIIT, running and all manner of hybrid workouts, I’ve been using the Rad R1 when I’m doing my strength and conditioning work in the gym like a good boy. They work for short runs and miles on the softer treadmill belt, while being stable and supportive enough to get under the bar and offering control for drills like box jumps and lunges. They look good, too.

New Balance Rebel V5 for $145, Adidas EVO SL for $105, Kiprun Kipride Max ($160). Another top-notch all-around shoe to rival the Saucony Endorphin Speed 5, the Rebel V5 is smooth, light and capable across the whole pace range. The Adidas EVO SL is a great alternative to the Saucony Endorphin Azura and can also handle anything you throw at it. But if you like your things super soft with a bit of bounce, the Kiprun Kipride Max serves up a cushioned plush ride with a bit of pop.

New Balance Fresh Foam X 1080 v15 for $170, HOKA Clifton 9 for $164: If you’ve never run before, the Hoka Clifton 9 is my recommendation for a beginner runner. Despite Hoka’s outsized (ahem) reputation, this is a pretty minimal shoe that’s comfortable, balanced, and light. —Adrienne So

Saucony Ride 17 for $110: This is also a good older budget-shoe model.

Saucony Hurricane 25 for $135, Brooks Glycerin 23 GTS for $180: Consumer tech director and podcast host Michael Calore runs in the Brooks Glycerin. This is our alternative pick if you’re shopping for shoes that offer greater stability.

Power up with unlimited access to WIRED. Get best-in-class reporting and exclusive subscriber content that’s too important to ignore. Subscribe Today.

Google’s upcoming Googlebook platform could launch with a much broader range of devices than many expected.

New findings suggest the first wave may include as many as eight laptops and tablets powered by chips from Intel, Qualcomm and MediaTek – giving buyers more choice from day one.

The discovery comes from newly uncovered development boards linked to Googlebook hardware. While Google has only teased that the first devices will arrive later this year, the latest evidence points to multiple manufacturers and several form factors too.

Chrome Unboxed reports that four of the devices appear to use Intel platforms, while another three rely on Snapdragon hardware. The publication also says that an additional device was built around MediaTek’s Kompanio Ultra processor, and could take the form of a tablet rather than a laptop.

The European Commission has appointed Jim Hagemann Snabe, chairman of Siemens’ supervisory board, as its special envoy for industrial artificial intelligence. He will advise Commission President Ursula von der Leyen and tech sovereignty chief Henna Virkkunen on how to accelerate AI adoption across European industry.

The backlash was immediate. Snabe’s appointment lands weeks after Siemens was among the companies that lobbied hardest for the rollback of the EU’s AI Act, the world’s most ambitious AI regulatory framework. Critics say the appointment amounts to handing advisory power over AI policy to the same industry that successfully weakened it.

Who is Jim Hagemann Snabe

Snabe, 60, is a Danish executive who co-led SAP as co-CEO from 2010 to 2014 before moving to the supervisory board. He became chairman of Siemens’ supervisory board in 2018. Beyond those roles, he has served on the advisory board of Google Cloud, on the board of US enterprise AI firm C3.ai, and as a board of trustees member at the World Economic Forum.

The Commission says it conducted a thorough conflict-of-interest assessment before the appointment. For the duration of his mandate, which runs until 31 March 2027, Snabe will suspend his Google Cloud and C3.ai memberships. The role is unpaid.

The AI Act rollback

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

The timing is what makes the appointment politically charged. On 7 May, the Council of the EU and the European Parliament reached a deal to simplify the AI Act through the so-called Digital Omnibus. The headline change was a 16-month delay to high-risk AI obligations, pushing the deadline from August 2026 to December 2027.

More significantly for Siemens, the deal introduced an industrial AI exemption. AI systems used on factory floors and embedded in machinery will now be covered by separate machinery regulations rather than the AI Act, unless a failure could directly endanger health or safety. Germany, where Siemens is headquartered, led the push for that exemption. Chancellor Friedrich Merz called for freeing industrial AI from the EU’s “regulatory straightjacket” at the Hannover Messe trade fair in April, with Siemens executives alongside him.

Virkkunen, who drove the simplification through the College of Commissioners, framed the deal as proof that Europe can maintain a rules-based approach while making regulation workable for industry. Snabe’s appointment is the next step in that direction: an explicit signal that industrial competitiveness, not regulatory caution, is now the priority.

The criticism

“My first reaction was just: Wow,” said Kim van Sparrentak, the Dutch Green lawmaker who led the Parliament’s work on the AI Act. “They fought hard against AI rules for themselves, they lobby against technological sovereignty, and now they get to decide how we are going to integrate AI.”

The concern is not only about Siemens. Snabe’s board positions at Google Cloud and C3.ai place him at the intersection of the three constituencies most directly affected by EU AI policy: European industry, US Big Tech, and the enterprise AI software market. Suspending board seats is not the same as severing ties, and critics argue that an unpaid advisory role with no formal accountability is precisely the kind of arrangement that makes revolving-door governance difficult to scrutinise.

The Commission has not disclosed the specific terms of Snabe’s conflict-of-interest assessment. It says one was carried out but has not published the methodology or findings, which makes the assurance hard to evaluate independently.

What the role involves

Snabe’s mandate is to advise on how Europe can boost industrial AI adoption, a priority that the Commission has elevated since the AI Act’s passage exposed a tension at the heart of European tech policy: the desire to regulate AI and the fear of falling behind the US and China in deploying it.

The appointment was announced alongside the Commission’s broader technology sovereignty blueprint, which includes the Cloud and AI Development Act, Chips Act 2.0, and new restrictions on US cloud providers handling sensitive European government data. Snabe’s role sits within that framework, theoretically bridging the gap between Brussels’ regulatory ambitions and the corporate reality of getting AI into European factories.

Whether a Siemens chairman is the right person to bridge that gap or simply the most obvious symptom of the gap itself is the question Brussels will be debating for the duration of his mandate.

Uber blew through its entire 2026 AI coding budget by April. Microsoft revoked its developers’ Claude Code licences six months after enabling them. One company reportedly ran up a $500 million Claude bill in a single month after forgetting to set usage limits. A Priceline employee told TechCrunch that a routine Cursor contract renewal came back four to five times more expensive.

The pattern is the same everywhere. Per-token prices have collapsed, but the push for autonomous AI agents has sent consumption through the roof. Companies that gorged themselves on all-you-can-eat subscriptions in early 2025 are now scrambling to understand where the money went, and whether any of it produced a return.

The paradox in numbers

GPT-4-equivalent performance now costs roughly $0.40 per million tokens, down from $20 per million in late 2022. That is a 98% reduction. Yet enterprise AI bills have risen by an estimated 320%, according to multiple industry analyses. The average enterprise AI budget has grown from $1.2 million per year in 2024 to $7 million in 2026.

The culprit is volume. Agentic AI tools released since November 2025, including Anthropic’s Claude Opus 4.5, OpenAI’s GPT-5.1, and Google’s Gemini 3 Pro, have multiplied token consumption per task. A simple linear workflow in 2023 cost about $0.04 per interaction. An orchestrated agentic system in 2026 costs roughly $1.20, about 30 times more. Individual engineers at Microsoft were reportedly spending between $500 and $2,000 a month on tokens before the licences were pulled.

Nicholas Arcolano, head of research at engineering management platform Jellyfish, told TechCrunch that per-developer consumption has risen roughly 18.6 times in nine months. Engineers who used the most tokens were about twice as productive as lighter users, but they spent 10 times the tokens to get there. “Whether extreme spend pays off comes down to the ultimate business value of shipped code, which most companies still can’t measure,” Arcolano said.

From tokenmaxxing to guardrails

“Six months ago, I would have a conversation with a customer and it would be all about ‘What can it do? Is it good enough?’” Alexander Embiricos, OpenAI’s head of enterprise, told TechCrunch. “Now the conversations are about, ‘We’re spending so much. What visibility do you have? What token controls do you have?’”

J.R. Storment, executive director of the FinOps Foundation, described the shift bluntly. “In April and May, I started hearing from companies: ‘Oh my god, we are 3x over our entire 2026 token budget and it’s only April.’ The whole conversation shifted from tokenmaxxing and ‘go fast’ to ‘we need guardrails, how do we control this?’”

Priceline’s senior director of IT finance, Chris Reed, drew a comparison to the telecom billing era. “It’s like the crack-cocaine epidemic. They let you try it to get you hooked, and now you’re kind of beholden to it.” The company has begun placing token limits on certain groups. Reed said he is already seeing discrepancies between vendor-reported usage and Priceline’s internal data.

The Tokenomics Foundation

It is against this backdrop that the Linux Foundation this week unveiled plans for the Tokenomics Foundation, a new standards body aiming to bring the same cost discipline to AI tokens that FinOps brought to cloud spending.

The Foundation plans to build a canonical definition of “tokenomics,” open standards for AI token usage and billing, and new metrics including cost-per-intelligence and tokens-per-watt. A formal launch is planned for July. Nishant Gupta, chief availability officer at Salesforce, said in a statement that “token economics is fundamentally more abstract and opaque than anything we’ve managed at this scale before.”

The challenge is enormous. “Tracking cloud costs is a hundreds-of-millions-of-rows-a-month data problem,” Storment said. “Tracking token costs is a trillions-of-rows-a-month data problem.”

A market forms around the problem

Startups and established vendors are racing to fill the gap. Pay-i tracks and optimises AI spending. Paid lets developers bill based on actual value rather than subscription fees. Jellyfish, Waydev, and Faros AI provide agent monitoring to prove the ROI of developer tools. Ramp has moved into AI spend management. Datadog and New Relic have added token-level observability.

Model routing is emerging as the primary cost lever. Factory, an enterprise AI coding startup, launched a model router this week that automatically picks the cheapest adequate model for each task. Vitaly Gordon, CEO of Faros AI, said frontier labs are already doing this internally. “The financial report for how much you spend on Anthropic, even if you call the Opus model, some of the spend will be on Sonnet or Haiku, because they are smart enough to do it,” he said.

Goldman Sachs projects global token usage will multiply 24 times by 2030. The companies already over budget need solutions now, and the Tokenomics Foundation’s first deliverable is still months away. As Gordon put it: “Maybe we created a steam engine, but we still haven’t figured out the assembly line.”

iFi Audio is back in the dongle DAC fight with the new GO link 2 Max, a compact USB-C DAC/headphone amplifier designed for smartphones, tablets, laptops, and PCs. Announced around High End Vienna 2026, the new model lands at $85 USD which puts it directly into one of the most crowded corners of personal audio.

And crowded is being polite.

The dongle DAC category is now packed with options from iFi, FiiO, Shanling, AudioQuest, Schiit Audio, Questyle, and enough other brands to make your phone’s USB-C port consider early retirement. AudioQuest has a new model coming as well, so clearly nobody got the memo that the boat was already full and starting to take on water.

Still, iFi has been at this long enough to know the assignment. The GO link 2 Max is not trying to be a desktop replacement, a battery-powered Bluetooth DAC, or a tiny slab of CNC-machined jewelry with a price tag that makes you clean your glasses, reload the page, and wonder if someone misplaced a decimal point. It is a wired USB-C dongle DAC with more output power, dual-DAC architecture, iFi’s S-Balanced technology, and app-based firmware support for under $100.

That might actually be a good deal if the sound quality has been improved and the cable can take the abuse.

Dual ESS Sabre DACs in a Tiny USB-C Package

The GO link 2 Max uses a dual ESS Sabre DAC architecture, with one DAC chip assigned to each audio channel. iFi says the design improves detail, definition, and instrument separation versus a single-DAC layout.

Format support is also strong for the price: PCM up to 32-bit/384kHz and native DSD256. That is more than enough for the overwhelming majority of users streaming from Qobuz, TIDAL, Apple Music, or a local hi-res library. Nobody needs to pretend they are casually commuting with 11.2MHz DSD files. Call your therapist if that’s actually something on your smartphone.

The GO link 2 Max also uses iFi’s GMT clock circuitry with a specialized crystal oscillator, along with ESS technologies such as Time Domain Jitter Eliminator. The goal is lower distortion, cleaner timing, and better clarity from a device small enough to disappear into a pocket.

More Power Than the Size Suggests

The headline number is up to 241mW of output power, which is a lot for something this small and affordable. That does not mean it will replace a proper desktop headphone amplifier, and nobody should expect it to drive planar headphones without some strain at higher levels.

But for IEMs, efficient dynamic headphones, and many portable over-ear models, 241mW gives the GO link 2 Max enough muscle to be more than a basic USB-C phone adapter with delusions of grandeur. Han Solo would understand.

In our review of the previous iFi GO link Max, the appeal was clear: it was small, solidly built, genuinely plug-and-play, and offered a lot more volume, resolution, clarity, bass texture, imaging, and separation than a basic laptop or phone headphone output.

It also brought dual ESS Sabre DACs, 32-bit/384kHz PCM, DSD256 support, and a 4.4mm balanced output to the sub-$100 category, which made the $79 price feel like someone at iFi had either lost a bet or found religion.

The limitations were also clear: the attached USB-C cable was a structural weak point, the 3.5mm output had less power than the 4.4mm jack, and high-impedance dynamic headphones were not always the best match.

The GO link 2 Max appears to stay focused on the same core idea, but with more output power, dual DACs, Dynamic Range Enhancement, THD compensation, and better software support through iFi Nexis.

That is the right direction.

S-Balanced Output, Not a 4.4mm Balanced Jack

One detail needs to be stated accurately: the GO link 2 Max does not appear to add a 4.4mm balanced headphone output. Instead, it uses iFi’s S-Balanced technology through its 3.5mm headphone output.

iFi says S-Balanced applies balanced circuit principles to a single-ended 3.5mm output to reduce channel crosstalk and improve separation. According to iFi, the implementation cuts crosstalk between channels in half.

That distinction matters because “balanced” gets thrown around in portable audio like free drink tickets at a trade show. This is not the same thing as a 4.4mm balanced output. It is iFi’s own approach to lowering noise and improving separation from a standard headphone jack.

For most users with 3.5mm headphones and IEMs, that is probably more useful than adding another cable standard to the drawer of shame.

Dynamic Range Enhancement and Lower Distortion

The GO link 2 Max also includes Dynamic Range Enhancement, or DRE, which iFi says adds up to 6dB of additional range between the quietest and loudest moments in the music.

iFi also claims its THD compensation reduces distortion by more than 50% compared to the original GO link Max. That is a useful claim, but again, the listening test matters. Measurements can tell part of the story. Headphones, IEM sensitivity, source device behavior, and volume control implementation will tell the rest.

iFi Nexis App Support, But Android Gets the Good Stuff

The GO link 2 Max supports the iFi Nexis app, which enables over-the-air firmware updates, selectable digital filters, and volume limiting.

There is a catch: iFi says those Nexis features are exclusive to Android devices. That means iPhone, iPad, and Mac users should not assume they are getting the same app-based control experience.

The two selectable digital filters are hybrid and linear, giving Android users some control over the DAC’s sonic behavior. Whether most listeners will hear a dramatic difference is another matter. Digital filters are useful, but they are not fairy dust. They tend to make subtle changes, not convert a $85 dongle into a $2,000 desktop DAC because someone tapped the right button.

Hardware Volume Control Is the Smart Move

One practical feature is the GO link 2 Max’s hardware-based volume control. iFi says this lets users adjust volume without reducing digital resolution in the way software volume control can.

That matters most with sensitive IEMs, where small volume changes and low noise are important. It is not the flashiest feature on the spec sheet, but it is the kind of detail that can make a portable DAC easier to live with every day.

Specifications Compared

The Bottom Line

The iFi GO link 2 Max is for listeners who want a real upgrade from a phone, tablet, or laptop headphone output without carrying a desktop DAC or another battery-powered box. For $85, it offers dual ESS Sabre DACs, up to 241mW of output, S-Balanced technology, hardware volume control, and hi-res PCM/DSD support in a tiny USB-C package.

The dongle DAC market is packed tighter than a CanJam elevator, but this one stands out by focusing on the basics: more power, cleaner conversion, and better control for IEMs and efficient headphones.

Where to buy: $85 at Crutchfield | iFi Audio

Summer Game Fest 2026 ended with a bang on Friday with the reveal of the final game of the Final Fantasy 7 Remake trilogy, Final Fantasy 7 Revelation. Not only did we get to finally see the trailer for the last entry, we were treated to gameplay footage as well. And we won’t have to wait long — it’s expected to come out next spring. 

The game picks up right after the previous game, Final Fantasy 7 Rebirth, and includes all the fan-favorite characters and ties up the loose ends of this retelling of the story from Final Fantasy 7. 

The trailer showcases the final chapter, in which Sephiroth wields incredible power and it’s up to Cloud and his allies to stop him. Featured in the video is the final character to join the team, Vincent Valentine, who appeared very late in the previous game, Final Fantasy 7 Rebirth. We also see the Weapons – Ruby, Emerald and Ultimate – that will be available in battle for those who want a challenge. Revelation also brings back the ability to swap characters on the fly during battles. 

The Summer Game Fest stream included a new addition to the game, Function Integrated Tactical Suitwear, or FITS. This system will unlock new movesets and battle boosts while allowing players to customize the look of their characters. 

The Final Fantasy 7 Remake series began in 2020 as a retelling of the original game from 1997.  The three games expand on the original’s story while modernizing the graphics and combat, resulting in a series that offers significantly more game to play.  

Final Fantasy 7 Revelation will launch in 2027 on PC, PS5, Xbox Series consoles and the Nintendo Switch 2. 

If you’re still using Office 2019 on your Mac, your time may be running out.

Microsoft has confirmed that from 13 July, Office 2019 for Mac will lose the ability to create, edit and save documents due to an expiring security certificate. While the apps themselves won’t suddenly disappear, they could become far less useful. This is especially problematic for anyone who still relies on Word, Excel or PowerPoint as part of their daily workflow.

The situation is unusual because Microsoft sold Office 2019 as a one-time purchase rather than a subscription. Many buyers picked it up specifically to avoid recurring Microsoft 365 fees and were happy to stick with a version that covered the basics without constantly adding new features.

Microsoft officially ended support for Office 2019 for Mac back in October 2023. Until now, that largely meant no new features or security updates, while the apps continued working as normal. The upcoming certificate expiry changes that.