from the bad-faith-bullshit dept

Paramount is clearly getting nervous about the growing opposition to its $111 billion merger with Warner Brothers, which is being intensely criticized for dodgy overseas funding, its dire impact on journalism, and the inevitable mass layoffs, consumer price hikes, and shittier overall product that always results from debt-fueled mega-media consolidation.

There’s a certain desperation creeping into their arguments as state regulators send signals that they’re considering filing an antitrust lawsuit. Top Paramount lawyer Makan Delrahim recently sat down for an interview with the billionaire-owned LA Times (non-paywalled alternative), and insisted that opposition to the company’s terrible merger spree is somehow antisemitic:

“Let’s be honest,” he told the Times. “There’s a lot of fear-mongering, particularly from people in Washington, D.C. They are running a political campaign. Some of these people are trying to inflict harm on this transaction, really because of their own antisemitic views. Regulators and law enforcement officials will see right through that.”

That is, of course, a whole lot of bullshit. Delrahim is trying to pretend that opposition to the deal stems from the fact that billionaire Trump-donor Larry Ellison, who has retooled CBS News to be more friendly to Benjamin Netanyahu, is Jewish. But if there’s any personal ire directed at Ellison as it pertains to the deal, it’s that he has a generational track record of being a foundationally terrible person.

The real-world concerns about the deal have focused on things like the fact it’s heavily financed by Saudi Arabia and China. And there’s fifty years of history showing that deals like this (especially deals involving Warner Brothers) routinely result in mass layoffs, higher prices, and both a shittier company and a less healthy film and television production market.

This sort of mindless consolidation is generally just a shell game performed by the extraction class and the kind of people obsessed with scale that have no genuine, original ideas. It’s utterly senseless, extractive, and destructive, as we all saw with the disastrous AT&T–>Discovery–>Warner Brothers mess (and the AOL Warner Brothers mess decades earlier).

Quick refresher: Delrahim was Trump’s DOJ “antitrust enforcer” during his first term. Delrahim “enforced antitrust” by doing things like rubber stamping Sprint’s merger with T-Mobile, which immediately resulted in more than 8,000 layoffs and an abrupt end to what passed as price competition in U.S. wireless.

These are, you’ll be surprised to learn, bad faith actors who aren’t actually interested in the public interest, product quality, happy workers, healthy markets, healthy companies, or much of anything else beyond short-term financial gains, tax breaks, control, and outsized higher-level executive compensation.

Ellison and Delrahim don’t have to worry about the Trump DOJ or FCC interfering in the deal. But their desperation suggests they are definitely nervous about negative public perception, European regulatory approval, and the hints being sent by state attorneys general that they’re cooking up a collaborative antitrust lawsuit that could either block or dramatically extend the project timeline.

Filed Under: antisemitism, antitrust, consoliation, journalism, larry ellison, makan delrahim, media, mergers

Companies: paramount, warner bros.

Microsoft has created an open-source fork of Windows Terminal called “Intelligent Terminal,” and it allows you to use AI directly inside Terminal without interfering with the regular session.

Microsoft describes the Intelligent Terminal as a built-in assistant that can help you explain errors, draft commands, and fix problems without leaving the terminal.

First, the agent can stay aware of what is happening in your terminal and help when a command fails. Second, it can remember active and past agent sessions, so you can return to earlier work without losing your place.

When you open Intelligent Terminal for the first time, it lets you choose the AI agent for the Terminal pane.

In my screenshot, it lists GitHub Copilot, Claude, Codex, and Gemini. GitHub Copilot is shown as “will be installed,” while the others are already installed.

There are also separate toggles for Automatic error detection and Automatic error suggestion.

When you turn on error detection, Terminal can notice failed commands. Similarly, error suggestion goes further and sends the error to the selected AI agent for a possible fix.

There’s another option, Session management, that lets Intelligent Terminal track active and past agent sessions. This is what allows you to reopen previous agent work.

Once you’ve configured Terminal AI, it’s quite easy to use. Terminal opens with an AI pane below the shell, where it says “Welcome to Intelligent Terminal.”

In my hands-on, I selected Claude as my Terminal AI model, which is why Claude Code is running inside the pane. It could plan a coding task and then ask whether I wanted to auto-accept edits, manually approve edits, or keep planning.

On the left side, you can choose to show or hide the agent panel and turn error detection on or off through its icon. On the right, you’ll see the agent management icon that opens your session management panel and agent status bar.

Intelligent Terminal’s Resume session is one of its best features

As a developer, I use Claude Code in Windows Terminal a lot for help, and while it does the job well, the only issue is that you can’t resume sessions in the standard Terminal unless you’re willing to use Claude’s built-in resume skill, which often makes the model perform worse.

Current Windows Terminal does have a toggle that allows it to open previously closed tabs, but that doesn’t restore your previous sessions.

Intelligent Terminal addresses these concerns with the ability to resume sessions, so you can always go back and forth between your earlier agent work.

Terminal AI is a great idea, but it’s not meant for everyone, and Microsoft understands that, which is why it’s a separate app, and it’s not included with Windows installations yet.

If you’re interested, you can download Intelligent Terminal from the Microsoft Store or Github.

Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.

The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.

Advertisement

Get the whitepaper

It’ll come as a free update for players who own the game already.

The battle to rework the failed Apple Intelligence initiative and Siri stems from a fateful meeting of executives that triggered a major restructuring of Apple’s AI efforts. Here’s what happened, and when it all went down.

Monday’s WWDC keynote address is expected to be heavy on AI features in iOS 27 and Apple’s other operating systems.

It should also bring to an end a turbulent period for the company. After the initial launch of Apple Intelligence and its seeming failure, as well as repeated delays for the promised Siri revamp, it had to do something.

Over the last two years, Apple had to make changes, which also included a considerable restructuring of its AI work. According to Mark Gurman in Sunday’s “Power On” newsletter for Bloomberg on Sunday, the overhaul effort stemmed from a meeting in early 2025.

A big AI meeting

At the time of the meeting, Apple was facing an industry that was moving extremely fast, leaving its own AI work behind. Executives met in a conference room near the software engineering department to try and solve the problem.

Though CEO Tim Cook wasn’t there, now-retired COO Jeff Williams was the one who called the meeting to order. Other executives in attendance include multiple C-level executives, as well as former interface design chief Alan Dye and Apple Vision Pro lead Mike Rockwell.

The meeting was all about the crisis that was Apple Intelligence and the looming prospect of a delayed Siri update. Executives quickly realized the scale of the problem, and its impact on Apple if changes didn’t happen soon.

Craig Federighi promoting Apple Intelligence – Image Credit: Apple

Advertisement

The meeting then moved to make a recommendation to Cook about Apple’s response. At the time, Cook had little confidence in then-AI chief John Giannandrea, also in attendance at the meeting.

Software chief Craig Federighi led most of the talks, but Rockwell volunteered to take the role to fix AI and Siri. Rockwell’s credibility was high, following the launch of the Vision Pro headset, which helped his cause.

A decade previously, former hardware head Dan Riccio raised the need for an AI leader to be on the Apple executive team. He also told Rockwell to make a five-year plan to rework Siri.

However, at the time, the top executives weren’t that receptive to the idea, and the Siri roadmap wasn’t completed. By the 2025 meeting, the same group of executives believed that there needed to be some leadership changes, and recommended to Cook for Rockwell to manage Siri.

Siri, not all AI

While Rockwell was recommended and Cook was close to approving the plan in March 2025, it wasn’t a done deal.

At the time, Rockwell thought he was working to become Apple’s AI leader in general, replacing Giannandrea. Federighi, however, believed that Rockwell should oversee Siri and report to him, instead of directly to Cook.

Rockwell considered that Federighi wasn’t seeing AI as being important, and then started to back away from the Siri role. With a yearning to become a senior vice president, Rockwell wanted promotion, but eventually agreed to the Siri position under Federighi.

Model behavior

That Siri role decision meant that Apple still had to find someone to deal with AI models, which led to a a long period of headhunting in 2025. Eventually, Amar Subramanya was picked to be the second AI leader, again reporting to Federighi.

Google Gemini is being used to help create Apple’s models.

However, Apple still had to catch up with the rest of the industry. To that end, Rockwell started to look at ways to do so, including using third-party solutions.

That eventually resulted in Rockwell, Federighi, and Eddy Cue making a deal with Google to use Gemini and Google Cloud to jumpstart creating the new Apple Foundation Models.

More Cook than usual

While the meeting was a big driver for AI, Apple’s seeming failure at the time also led to more input from Cook. At the time of the meeting, Cook decided to inject himself into work on the AI roadmap, making more decisions about plans, and even delivered an AI pep talk to the company.

Cook became a lot more hands-on with AI than he normally would for company projects. Roadmap and key decisions were previously left to his reports, with Cook usually taking a light-touch approach to management, but not for AI.

He urged Federighi and others to treat AI more seriously and to make it a success.

Federighi, in charge of implementing AI features, has adjusted his view and handling of the technology. He now views it as the central focus of operating system upgrades for years.

That early 2025 meeting was a turning point for Apple’s AI work, and came at a crucial time for the company. We won’t know if it has done enough until the Keynote video begins on Monday.

FiiO came to High End Vienna 2026 with three new desktop hi-fi products, and the most traffic-friendly one is almost certainly the Jade Audio LEVEL 1 desktop amplifier. Why? Because FiiO is claiming up to 2 x 300 watts of output from a compact desktop amp priced at $139.

That is either one of the more interesting budget hi-fi announcements of the show, or someone at FiiO found the “make forum threads explode” button and pressed it with both thumbs. Probably a little of both. Hopefully it came with schnitzel.

The new lineup also includes the FiiO JT9 open-back planar magnetic headphones and a fully discrete Class A headphone amplifier, giving FiiO a broader desktop ecosystem aimed at listeners who want more than another dongle DAC dangling from a laptop.

FiiO LEVEL 1: A $139 Desktop Amp With Real Ambition

The FiiO Jade Audio LEVEL 1 is a compact desktop stereo amplifier built around the Texas Instruments TPA3255 amplifier chip, which uses TI’s PurePath Ultra-HD technology. FiiO claims output up to 2 x 300 watts, but that number needs proper context: the headline rating is into 4 ohms at under 10% THD+N with a 53V/12A power supply.

That does not make the figure fake, but it does mean readers should not assume the LEVEL 1 will deliver 300 clean watts per channel into every speaker under normal conditions while also making espresso. FiiO’s own specifications also list 240W + 240W into 4 ohms and 135W + 135W into 8 ohms at under 1% THD+N, which are still very serious numbers for a compact amplifier at this price.

Can we also retire “giant killer” while we’re here? It gets tossed around far too often in hi-fi, usually before anyone has measured the thing, listened to it properly, or explained what giant was allegedly slain. And we’re not referring to some French kid playing basketball in Texas.

Connectivity Is the Real Story

Connectivity is stronger than expected for the money. The LEVEL 1 includes RCA line input, USB DAC input, coaxial digital input, and Bluetooth 6.0 reception with SBC and LDAC codec support. FiiO has also included a single-ended line output, subwoofer output, speaker outputs, physical bass and treble controls, and support through the FiiO Control app.

The USB DAC input supports up to 96kHz/24-bit playback, while the coaxial input supports up to 192kHz/24-bit. That gives the LEVEL 1 enough flexibility for desktop systems, small-room loudspeaker setups, CD transports, streamers, computers, and Bluetooth playback from a phone or tablet.

That matters because this is not being pitched as a minimalist purist amplifier. This is a compact, flexible desktop or small-room amplifier for listeners who want wired digital input, analog input, wireless playback, tone controls, subwoofer output, and enough claimed power to make people start arguing on Head-Fi and Audio Science Review before mother makes their breakfast.

Retro Styling, Modern Inputs, Very FiiO Pricing

The LEVEL 1 uses an aluminum alloy chassis with wood side panels, giving it the retro desktop look FiiO clearly wants. FiiO says the amplifier uses a six-layer immersion gold thick copper PCB and capacitors sourced from Germany and Japan.

The unit measures about 181 x 133 x 36mm, excluding feet, and weighs about 676g. It is available in black with walnut side panels or silver with maple side panels.

At $139, the LEVEL 1 lands directly in the value-fi combat zone where Fosi Audio, SMSL, Topping, Loxjie, WiiM, and Dayton Audio have been making life uncomfortable for traditional entry-level integrated amplifiers.

The big question is not whether the LEVEL 1 is inexpensive. It is. The question is whether it sounds composed, stable, and clean enough at real-world listening levels to be more than another spec-sheet stunt that ends up in the hi-fi dustbin of history.

FiiO JT9: Open-Back Planar Headphones Join the Desktop Push

FiiO also used High End Vienna 2026 to preview the JT9, an open-back planar magnetic headphone built around a custom 95 x 86mm planar driver and ultra-thin diaphragm. FiiO says the JT9 uses its dual-coating driver technology and a uniform magnetic field design, with the goal of improving transient response and tonal accuracy.

The JT9 is also designed to be more portable than its large planar driver might suggest. Current show information lists the headphones at a claimed 365 grams, with a foldable design for easier storage. Sensitivity is listed at 95dB/mW, and FiiO is including both 3.5mm single-ended and 4.4mm balanced cables.

Pricing has not been confirmed yet, and that matters. The JT9 enters a very crowded planar headphone category where FiiO already competes with models like the FT1 Pro and JT7, while brands such as HiFiMAN, Audeze, Moondrop, Sendy Audio, and others have made life rather interesting at multiple price points. If FiiO prices the JT9 aggressively, it could become one of the more interesting open-back planar headphones to watch. If not, it becomes another planar headphone looking for room at the Waffle House counter and we know how that usually ends up at 3 in the morning.

FiiO CLASS A Headphone Amplifier Rounds Out the Desktop System

The third product in FiiO’s High End Vienna 2026 desktop lineup is the FiiO CLASS A, which appears to be the actual product name. Subtle? Not exactly. Accurate? Apparently.

Current show information describes the CLASS A as a fully discrete pure Class A headphone amplifier rated at 1000mW + 1000mW. FiiO is also listing a 60W toroidal transformer, discrete regulated power supply, five selectable gain settings, bass and treble controls, and 12V trigger input/output.

Connectivity includes 3.5mm, 4.4mm balanced, and XLR headphone outputs, along with RCA input and RCA output connections.

The CLASS A also does not appear to include a built-in DAC, which fits the concept. This is being positioned as a standalone headphone amplifier for listeners who already have a DAC or desktop source and want FiiO’s take on pure Class A amplification, complete with the usual trade-offs: more heat, lower efficiency, and hopefully the smoother, richer presentation people expect from this type of circuit.

That suggests FiiO is not just throwing one inexpensive desktop amp at the wall. The company is building out a broader desktop ecosystem aimed at both loudspeaker and headphone listeners. Between the Jade Audio LEVEL 1, JT9 planar headphones, and CLASS A headphone amplifier, FiiO clearly wants more desk space. Given the pricing history, it may get it.

The Bottom Line

The FiiO Jade Audio LEVEL 1 is the key product here: a $139 desktop amplifier with TPA3255 amplification, USB DAC, coaxial, RCA, Bluetooth 6.0 with LDAC, tone controls, pre-out, subwoofer output, and high claimed power output that needs independent testing. Alongside the JT9 planar headphones and FiiO CLASS A headphone amplifier, it shows FiiO expanding its desktop hi-fi lineup beyond portable DACs and headphone amps.

Readers should care because the LEVEL 1 offers an unusual mix of power, inputs, outputs, and price. If it performs well in real systems, it could be a useful option for desktop setups, small rooms, powered subwoofer integration, and affordable two-channel systems.

Where to buy:

Related Reading:

The 2026 World Cup kicks off next week across 16 cities in the US, Mexico, and Canada. It is the largest in history: 48 teams, 104 matches, 39 days. It is also the most technologically surveilled sporting event ever staged, with robot dogs, net-shooting hunter drones, and thousands of AI-powered cameras deployed across venues and fan zones.

“It’s 78 Super Bowls over 39 days,” said Andrew Giuliani, executive director of Trump’s World Cup task force. FEMA has distributed $625 million to the 11 US host cities, with an additional $250 million earmarked for tracking and neutralising suspect drones.

Boston Dynamics’ Spot robot dogs are patrolling AT&T Stadium in Arlington, Texas, as part of a “Security Spot” initiative by owner Hyundai. The company says it is deploying “its largest and most advanced mobility fleet to date,” making it the first official partner to provide robotics for the tournament. The robots will inspect suspicious packages and hazardous materials.

Sightings of the quadrupeds sparked immediate surveillance fears. Rumours spread on social media that the dogs were scanning faces. Boston Dynamics told Chron that the robots “do not have facial recognition capabilities.” But the optics of robotic security patrols at a sporting event drew comparisons to the Black Mirror episode “Metalhead.”

Mexico is deploying four robot dogs called K9-X across its three World Cup venues. Officials told Wired the robots will intervene in fights or drunken incidents to protect officers. Technical details and the manufacturer were not disclosed.

Drones are the top concern. “If there is one threat that keeps me up at night, it is from drones,” said New York City Police Commissioner Jessica Tisch. Drones are banned over stadiums and fan zones. Counter-drone technology has become one of the fastest-growing segments in defence tech, and the FBI says it has a “full suite of options” to thwart incursions.

One of those options comes from Fortem, which claims to have signed a multimillion-dollar DHS contract for quadcopters that shoot nets at encroaching drones to trap them in midair. DHS declined to discuss the contract.

AI-generated misinformation is another concern. Officials warned that state actors could use deepfake videos to sow panic. The FBI will activate joint operations centres in each host city on match days. “If there’s a video that shows an explosion going off at a site, and it’s AI generated, we have people on the ground who can validate whether or not that’s true,” said FBI Special Agent in Charge Amit Kachhia-Patel.

In Dallas, a $120 million tech upgrade gives police body cameras with real-time translation, helping officers communicate with international visitors. Several drone detection startups are also joining federal efforts to secure the skies.

The security apparatus comes against a tense backdrop. The US-Israel war with Iran has changed the threat picture since planning began. Human rights groups have raised concerns about ICE presence at venues. And the Secret Service, in charge of protecting world leaders who attend, is understaffed by about 860 agents, according to Homeland Security Secretary Markwayne Mullin.

“We’re as prepared as we can be,” Giuliani said. The unspoken bet is that the tech will make up the difference.

OpenAI has begun rolling out Lockdown Mode to ChatGPT, a new security setting designed to block attackers from stealing data through prompt injection attacks. The feature disables live web browsing, agent mode, deep research, image retrieval, Canvas networking, and file downloads. It is available to logged-in users across Free, Go, Plus, Pro, and self-serve ChatGPT Business plans.

Prompt injection remains what OpenAI calls a “frontier” problem affecting all large language models. The attack works by hiding malicious instructions in content the model processes, such as a webpage or uploaded file. If the model follows those instructions, it can be tricked into sending sensitive data to an attacker-controlled server.

Lockdown Mode does not stop injections from happening. A malicious payload embedded in a cached webpage or uploaded PDF can still influence the model’s behaviour. What it does is shut down the outbound pathways an attacker would use to exfiltrate the data. No live browsing means no network requests to external servers. No image retrieval means no pixel-based data channels.

“Lockdown Mode is designed to substantially reduce the risk of prompt injection-based data exfiltration, but it does not guarantee that data exfiltration cannot happen,” OpenAI said. “Risk may remain through enabled Apps, unforeseen combinations of capabilities, or newly discovered techniques.”

The trade-off is significant. With Lockdown Mode on, ChatGPT loses most of what makes its agent and research features useful. Live browsing drops to cached content only. Agent mode is gone entirely. Deep research is disabled. It is, as OpenAI acknowledges, “not intended for everyone.”

The feature arrives as prompt injection attacks on AI agents have become a growing concern. Security researchers have demonstrated hijacks against agents from Anthropic, Google, and Microsoft via their GitHub Actions integrations. All three paid bug bounties but published no public advisories. The underlying weakness is fundamental: LLMs cannot reliably separate data from instructions.

Lockdown Mode and Developer Mode cannot be used simultaneously. Turning one on disables the other. OpenAI also launched a separate session management feature that lets users review active ChatGPT sessions and log out of individual devices if they spot unauthorised activity.

The feature is a pragmatic concession. OpenAI is not claiming to have solved prompt injection. It is accepting that the problem persists and offering users a way to reduce their exposure by giving up functionality. For anyone handling sensitive data in ChatGPT, that trade-off is worth making. For everyone else, the expanding agent ecosystem and its growing attack surface mean the risk is only increasing.

A data breach at the dental benefits administrator DentaQuest has reportedly exposed the sensitive data of 2.6 million accounts.

The security incident came to light last month, when the infamous extortion group ShinyHunters listed the company on its data leak site and claimed to have stolen more than 234 GB of data.

Following what the threat actor describes as a failure to reach an agreement with the company, the data was publicly leaked.

DentaQuest, part of Sun Life, is one of the largest dental benefits administrators in the United States. It manages dental insurance plans and provider networks for Medicaid programs, Medicare Advantage plans, employers, health plans, and individual customers.

The company says it serves 35 million customers, operates programs in 50 states, and has a network of 140,000 dentists and dental specialists.

On June 2, DentaQuest confirmed on its website that its networks had been breached and the incident caused “limited disruption” in customer service.

“DentaQuest is actively managing a cybersecurity incident involving unauthorized access to a limited portion of our network,” reads the statement.

“Upon discovery of the initial incident, we took immediate action to secure our environment, contain the attack, and mitigate the threat.”

“Our systems remain fully operational, and we continue to serve our clients with limited disruption.”

The firm also stated that it engaged external experts to help with the investigation and determine the data that was compromised.

Yesterday, data breach alerting service Have I Been Pwned (HIBP) analyzed the leaked information and found that it contained records for 2.6 million accounts. Specifically, the following was exposed in the leaked dataset:

• Email addresses
• Full names
• Phone numbers
• Government-issued IDs
• Health insurance information
• Genders
• Dates of birth

Although DentaQuest’s statement did not confirm that the data breach affected its clients, HIBP is known to validate leaked datasets using multiple verification methods.

HIBP also stated that roughly 66% of the exposed records were present in its database from past incidents affecting other organizations and services.

People who may have had their information exposed in this incident should be cautious about all incoming communications, as the leaked data increases the risk of social engineering and phishing attacks.

Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.

The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.

Get the whitepaper