The AI industry has fully entered the “agent era,” a paradigm where AI models do far more than generate text — they now actively plan, execute, and course-correct complex tasks over days rather than seconds.

Thus, it’s perhaps unsurprising to see Chinese e-commerce giant Alibaba’s famed Qwen Team of AI researchers release a model capable of performing autonomous agentic AI work over multiple days: that model has arrived in the form of Qwen3.7-Max which the company reports in a blog post achieved “~35 hours of continuous autonomous execution” — albeit, in a proprietary, not open source format, as prior Qwen Team releases were.

This is also to be expected — it’s what many analysts and industry experts feared in the wake of the departure of several key Qwen Team leaders earlier this year. But it makes sense for Alibaba financially, at least in the short term: training AI models, especially ones as powerful as Qwen3.7-Max, is expensive, and giving them away essentially for free, as open source models are, does not immediately help recoup any costs.

In that sense, Alibaba is simply aligning its efforts with American AI giants like OpenAI and Google by offering the latest and greatest models only through paid APIs and subscription or paid web plan bundles, and slightly less performant ones through open source.

Still, the arrival of Qwen3.7-Max offers further optionality to enterprises and individual users, and more competition for American AI labs — rarely a bad thing for consumers at all budget levels. Yet, the fact that the model is only accessible from Chinese-based endpoints means it may be limited in its appeal to American and European enterprises seeking to maximize compliance and security posturing when fulfilling government contracts, or even just attempting to comply with all relevant state, local, and national data sovereignty regulations.

The marathon AI era

To understand why Qwen3.7-Max is a departure from previous models, one must look at how it was trained and how it operates in practice.

Language models typically degrade when forced to maintain a single train of thought over thousands of conversational turns; they forget instructions, hallucinate variables, or simply get stuck in logical loops. Qwen3.7-Max was specifically designed as a “versatile agent foundation” capable of “long-horizon reasoning” to overcome this exact bottleneck.

The starkest demonstration of this capability is an autonomous engineering task detailed by the Qwen team. The model was given access to an isolated server equipped with a T-Head ZW-M890 PPU—a hardware architecture the model had never encountered during its training. Its task was to optimize an attention kernel.

Over the course of 35 straight hours, Qwen3.7-Max operated entirely autonomously. It executed 1,158 distinct tool calls, performed 432 kernel evaluations, diagnosed compilation failures, and iteratively improved the code to achieve a 10.0x geometric mean speedup.

By comparison, Chinese competitor models like z.ai’s GLM-5.1 and Moonshot’s Kimi K2.6 capped out at 7.3x and 5.0x speedups respectively, often voluntarily terminating their sessions when they failed to make progress. However, both are available open source.

This endurance is achieved through what Alibaba calls “environment scaling”. Just as early LLMs grew smarter by ingesting more diverse text, Qwen3.7-Max was trained across a vast, scaled array of dynamic agentic environments.

It is capable of simulating a one-year lifecycle of a startup in the “YC-Bench” evaluation, navigating hundreds of decision-making rounds encompassing personnel management and contract screening. In this simulation, the model managed to generate $2.08 million in virtual revenue, nearly doubling the performance of the prior generation, Qwen3.6-Plus.

Furthermore, the model has built-in reward-hacking self-monitoring, autonomously detecting when it attempts to cheat a training environment and adding heuristic rules to correct its own behavior.

A brain for any scaffold

From a product perspective, Qwen3.7-Max is designed to be the cognitive engine for modern software development and enterprise automation.

The model offers a massive 1-million-token context window and a 64K maximum output limit, providing immense overhead for processing sprawling codebases or lengthy technical documents.

One of its most compelling features is “cross-harness generalization”. Rather than being hardcoded to work best within a specific proprietary interface, Qwen3.7-Max is built to act as a drop-in intelligence layer for diverse agent frameworks. It supports the Anthropic API protocol natively, allowing developers to plug it directly into existing tools like Claude Code or OpenClaw.

The benchmark data provided by Alibaba indicates that this generalized approach has paid massive dividends.

On the Apex Math Reasoning benchmark, Qwen3.7-Max scored 44.5, eclipsing Claude Opus-4.6 Max’s score of 34.5 and DeepSeek V4-Pro Max’s 38.3. It also posted dominant scores on Humanity’s Last Exam (41.4) and the realistic coding agent benchmark MCP-Atlas (76.4).

This translates into tangible utility for end-users. Through open source Model Context Protocol (MCP) integrations, the model can operate as an autonomous office assistant, capable of reading university formatting specs and automatically reformatting a messy Word document via command-line tools without human intervention.

Running this level of intelligence comes at a distinct cost. Developers accessing the API via Alibaba Cloud Model Studio will pay $2.50 per 1 million input tokens and $7.50 per 1 million output tokens. The platform also features explicit cache creation and read pricing, as well as a $10 fee per 1,000 calls for integrated web searches, though code interpreter tools remain free for a limited time.

Qwen3.7-Max occupies a strategic middle ground in the current API economy. While it demands a notable premium over aggressively priced domestic rivals—costing nearly double DeepSeek V4 Pro ($5.22) and Z.ai’s GLM-5.1 ($5.80)—it drastically undercuts the Western frontier giants it routinely matches on benchmarks.

For context, running heavy agentic workflows through OpenAI’s GPT-5.4 or Anthropic’s Claude Opus 4.7 will run developers $17.50 and $30.00 per million tokens, respectively. See VentureBeat’s pricing chart below:

By positioning Qwen3.7-Max just below Google’s Gemini 3.5 Flash ($10.50) but well above budget-tier models, Alibaba is signaling that this isn’t a commodity release; it’s a flagship reasoning engine priced to lure enterprise workloads away from Silicon Valley’s most expensive offerings.

Licensing remains proprietary for now

For all its technical brilliance, the most controversial aspect of Qwen3.7-Max is how it is distributed. Qwen is billing the release as a “proprietary model”. It is strictly API-only.

Historically, Alibaba’s Qwen has been a hero to the open-source and local LLM communities. Previous iterations, like Qwen 2.5 and Qwen 3.6, released their weights publicly. Open weights allow developers, researchers, and enterprises to download the model, run it on their own hardware, and fine-tune it for highly specific or data-sensitive use cases without sending proprietary information to a third-party server.

By locking Qwen3.7-Max behind an API, Alibaba is pivoting to the standard commercial playbook utilized by OpenAI (with GPT-4) and Anthropic (with Claude). For enterprise users, this means utilizing Qwen3.7-Max requires trusting Alibaba Cloud with their data streams and relying entirely on internet connectivity to run their agentic workflows. For the open-source community, it means losing access to what is currently one of the most capable models on the planet.

Community reactions split between awe and disappointment

The reaction from the developer community has been swift, characterized by a mix of profound respect for the engineering achievement and frustration over the licensing model.

Prominent AI commentator Sudo su (@sudoingX) captured the prevailing sentiment on X (formerly Twitter). “qwen is unreal,” they wrote. “they just dropped 3.7 max and it is beating opus 4.6 max on most of the benchmarks they ran”.

The technical metrics, particularly the model’s endurance, have left many in the field stunned. “the apex math number, 44.5 against opus 34.5, that is not a small gap,” Sudo su noted. “the 35 hours straight on a kernel optimization task with 1000+ tool calls is the part i keep rereading. that is the agent era thing actually happening, not a slide”.

The speed of Alibaba’s iteration is also drawing notice. With Qwen 3.6 released just last month, the leap to 3.7-Max highlights a relentless development cadence. As Sudo su observed, “nobody else is moving like this”.

Yet, the praise is heavily caveated by the shift to a closed ecosystem. The loss of the model weights is seen as a blow to the localized AI movement, which relies on state-of-the-art open models to push the boundaries of what can be done on consumer hardware or private enterprise clusters.

“one thing though, please open source this one too,” Sudo su pleaded in their post. “3.6 dense made the entire local llm ecosystem better. the max tier going api only would close a door we have been keeping open. give us the weights eventually”.

Qwen3.7-Max proves that the autonomous agent era is no longer a theoretical projection; it is a present reality capable of executing complex engineering feats while humans sleep. The only question now is whether this new frontier of AI will be a democratized resource you can download to your laptop, or an intelligence utility rented strictly from the cloud. For now, with Qwen3.7-Max, it is undeniably the latter.

from the ctrl-alt-speech dept

Ctrl-Alt-Speech is a weekly podcast about the latest news in online speech, from Mike Masnick and Everything in Moderation‘s Ben Whitelaw.

Subscribe now on Apple Podcasts, Overcast, Spotify, Pocket Casts, YouTube, or your podcast app of choice — or go straight to the RSS feed.

In this week’s roundup of the latest news in online speech, content moderation and internet regulation, In this week’s roundup of the latest news in online speech, content moderation and internet regulation, Mike is joined by civil liberties lawyer Jennifer Granick. Together they discuss:

Support the podcast by joining our Patreon, with special founder membership available until May 28th.

Filed Under: africa, ai, artificial intelligence, content moderation, deepfakes, trust and safety

Companies: apple, kickstarter, twitter

The Supreme Court could now weigh in on the Apple versus Epic case where Apple was found in contempt of an injunction and forced to allow all developers to link externally without commission.

The Apple versus Epic saga is nowhere near an end even if Epic is celebrating a victory prematurely. Even as the case returns to Circuit Courts, Apple is requesting the Supreme Court to review two specific issues it has with the proceedings so far.

In the Supreme Court filing viewed by AppleInsider, Apple shares that the scope of the anti-steering injunction exceeds the District Court’s limits set by CASA. It also argues that the injunction violation was issued in error due to suggesting it was violating the “spirit” of the law rather than the letter.

Its arguments in the 34-page filing suggest that the Supreme Court should take up these matters because Apple’s is a perfect vehicle to address these issues. Apple asserts that providing a decision would settle matters for future cases, and if left untouched, could cause the CASA verdict to be a dead letter.

Basically, Apple hopes that there are enough discrepancies to ensure the Supreme Court at least picks up the case. In the meantime, Apple will continue its proceedings with Epic in the lower courts.

The story so far

Epic sued Apple in 2020 on antitrust grounds, but Epic lost on every count except one. That count pertained to Apple’s anti-steering practices.

Epic’s ‘1984’ parody ad

Apple removed the anti-steering provisions and provided a new, if complex, way for developers to link to external purchases. It meant developers still owed Apple a commission, 12% or 27%, even if it directed customers to the web.

Even though Epic filed the case and it wasn’t a class action, the injunction was applied to all developers based in the United States. Apple clearly planned to appeal that point even then, but then things were made more complicated.

Epic filed a complaint, which resulted in Apple being found in contempt. However, the original injunction didn’t mention anything about Apple’s commission, and the violation was argued in spirit.

Various appeals and arguments later, and Apple has been told it is owed a commission, even on external links. The problem is, Apple would have to come back to court and decide on the commission rate with Epic.

That’s how the case has arrived at the Supreme Court. And even though Apple tried to get the proceedings halted in the lower courts, twice, it must now face both at once.

Apple’s arguments

The foundations of Apple’s arguments appear to be sound. The courts do appear to be ignoring the precedent set by CASA.

Epic’s iPod-like ad

The Supreme Court ruled that lower courts were exceeding their jurisdiction by applying injunctions outside the scope of a case. However, the 9th Circuit has argued that there is an antitrust exception to CASA that would allow the decision in Apple’s case to stand.

Apple believes very strongly that this effectively bypasses the Supreme Court’s ruling and authority. That’s why it said it would render the CASA case a dead letter.

The other argument also has to do with how the 9th Circuit does business. Apple argues that in the other Circuit Courts, civil contempt is applied only if the letter of the law is violated, not the spirit.

Even if you don’t care about any of this legal back and forth, it is still incredible that the Epic Games lawsuit has reached this point. It started with a “1984” parody ad starring an apple wearing sunglasses and could finish with setting incredibly important precedent via the Supreme Court.

If Apple wins the “in spirit” portion of its arguments, Apple gets to carry on with its previous 12% and 27% commission rates for external linking. It would also mean proceedings in the lower courts would return to appeals stages.

If the universal nature of the injunction is thrown out, then only Epic will be affected by Apple’s move away from anti-steering practices. It would mean a total and abject failure of a case that cost Epic over a billion dollars already.

Apple has requested that its petition be considered during the Supreme Court’s June 25 conference. Perhaps Epic’s CEO should hold off on celebrations until after that date.

You know your Google API key has leaked so you rush to disable it before bad actors can start running up charges on your account. Bad news: According to security researchers at Aikido, people can use the API keys for up to 23 minutes after a user deletes them, creating a window of opportunity that, when combined with Google’s automatic billing tier upgrades, can devastate victims.

“We’ve identified a substantial window where an attacker with access to a leaked Google API key can continue to misuse that credential, after the user believes the key is revoked,” Joseph Leon, a security researcher with Aikido, told The Register. “In that window, an attacker could run up charges, pull sensitive files uploaded to Gemini, and exfiltrate cached context.” 

Aikido tested the gap during 10 trials over two days. In each trial, researchers created an API key, deleted it, and then sent three to five authenticated requests per second until no valid response came back for several minutes. 

From the time a user deletes the Google API key to when it can no longer be used propagates gradually across Google’s infrastructure, he said. Some servers reject the key within seconds while others keep accepting it for 23 minutes.

What this means is that an attacker holding a deleted key can repeatedly send requests until one reaches a server that has not caught up, Leon said. If Gemini is enabled on the project, they can dump files that were uploaded and exfiltrate cached conversations.

The paper cited a similar problem researchers disclosed in December involving AWS keys. In that case, after deletion, attackers had a four-second window to exploit, and researchers showed how they could create new credentials in that time. 

“Four seconds was enough to matter on AWS,” Leon wrote in the paper. “Given recent attention to Google API keys used to access Gemini, we set out to measure how long Google’s API key revocation window remains open.” 

Flaws can hit devs with huge surprise bills

The Register has reported numerous cases of Google API key abuse in which developers are suddenly hit with five figure bills after their credentials are compromised. 

The problem was compounded in April after Google reworked its billing policy to include spending tiers for users. While developers initially thought of it as a way to limit costs, Google automatically upgrades that spending tier to the next highest level without their knowledge. 

For users who have been working with Google for more than 30 days and have spent more than $1,000 over the lifetime of the account, their cap can be increased from $250 to $100,000 if their usage spikes – a windfall for crooks if the credentials fall into the wrong hands.

Developers whose Google API keys were stolen told The Register that their bills rocketed up to five figures minutes after their credentials were stolen, as bad actors loaded up on Google’s Gemini models such as Nano Banana and its video production model Veo 3. 

Google issued refunds in the three instances that The Register brought to its attention, returning $154,000 to those developers.

The victims told The Register that, during the attack, they were frantically trying to shut down the spending and turn off access to their projects even as costs climbed by thousands of dollars. Leon said in cases where a Google developer tries to shut off access to their account, deleting the API key will still give crooks time to inflict damage. 

“It’s hard to put a dollar figure on it,” Leon told us. “The window averaged 16 minutes in our testing and stretched to nearly 23 at the worst. During that window, the success rate is wildly unpredictable. We saw minutes where over 90% of requests still authenticated, and others where fewer than 1% did. An attacker who knows this can send requests at high volume to maximize their odds of hitting a server that hasn’t caught up. For Google API keys with Gemini access, the damage isn’t just a compute bill. It’s the files and cached context an attacker can exfiltrate before the key actually dies.” 

Using VMs, Aikido tested its findings across three Google Cloud regions – east coast US, western Europe, and southeast Asia – then they spot checked those results on different dates. For each trial, Aikido deleted a single API key and sent requests from each of the three VMs in parallel, Leon wrote in the paper. 

“VMs further from the US picked up the deletion faster, which is the opposite of what you’d expect. We can’t say exactly why from the outside. Google’s request routing is more complex than ‘VM region equals server region,’ and a VM in Singapore isn’t necessarily talking to servers in Singapore,” the paper states. “But the pattern was consistent across trials, which points to something about regional infrastructure, caching, or routing affinity driving the difference.” 

The trial used keys with access to Gemini, but he observed the same behavior with keys scoped to other GCP APIs, such as BigQuery and Maps. Google has built faster revocation for other credential types, Leon said. 

He said Google’s service account API credential revocations propagate in about 5 seconds. Gemini’s newer API key format – the one that starts with AQ – propagates in about a minute. 

“Both run at Google scale. Both suggest this is technically solvable for Google API keys, too,” Leon wrote. 

But Google told Aikido it has no plans to address the 23-minute gap researchers found with its other API keys. 

“After reviewing our report, they closed it as ‘Won’t Fix (Infeasible)’ with the comment ‘the delay due to propagation of the deletion of these keys is working as intended,’ “ Leon told us.

The Register has reached out to Google about this research, but has not yet received a response.  ®

Samsung released its Odyssey G8 monitor this week, and the 32-inch display brings a full 6K resolution to gaming for the first time. Gamers who chase the sharpest possible picture now have a genuine new choice that delivers 6144 pixels across and 3456 pixels down on an IPS panel. That pixel count produces a density around 224 pixels per inch, which makes fine details in environments and character models stand out in ways lower resolutions simply cannot match.

Resolution alone doesn’t tell the entire story, as Samsung has sneaked something quite clever into the Odyssey G8, in the shape of a dual mode feature. Users can switch to a roughly 3K resolution and receive a refresh rate of 330 Hz instead of the standard 165 Hz at full 6K. Anyone who prioritizes ultra fluid motion over pixel-perfect clarity should switch between the two settings depending on the game or what your system is capable of at the time. Thanks to AMD FreeSync Premium and NVIDIA G-Sync compatibility, both modes will keep screen tearing at bay regardless of resolution.

Sale

Samsung 34′ Odyssey G5 Ultra-Wide Gaming Monitor with 1000R Curved Screen, 165Hz, 1ms, FreeSync Premium…

• OPTIMAL CURVE, TRUE IMMERSION: The all-encompassing 1000R display fills every part of your peripheral vision and draws you right into the character’s…
• WQHD RESOLUTION: The ultrawide WQHD monitor provides a seamless dual-monitor experience with a 1000R curvature that matches the human field of view…
• SMOOTH TRANSITIONS: This gaming monitor features a 165Hz refresh rate and 1ms (MPRT) response time, eliminating motion blur for a smooth, competitive…

Connectivity-wise, the Odyssey G8 is designed to meet the high needs of that screen, with one top-tier DisplayPort 2.1 connection that can manage the full bandwidth without sacrificing clarity. Two HDMI 2.1 connections are available for console owners and those who want to connect numerous devices at once. The accompanying stand adjusts in almost every direction you might want, including height, tilt, swivel, and pivot, ensuring that your monitor sits perfectly for long gaming sessions. A tiny RGB light on the back provides additional lighting options without taking the eye away from the screen.

The Odyssey G8’s color performance is refreshingly realistic rather than simply trying to put up some impressive numbers, as it covers 99 percent of the sRGB region and has a 1000-to-1 contrast ratio, which is typical for an IPS screen. HDR10+ Gaming compatibility is also included, which means that brightness and contrast levels can be adjusted scene by scene in compatible titles, which is especially useful when you’re in the thick of some fast-paced action. Brightness levels are normally enough for most rooms, but do not expect to set any records.

Driving a 6K display at a playable speed necessitates some serious hardware, as recent high-end graphics cards paired with tools like DLSS are just about up to the task in modern games, but for many users, the greatest benefit will come from playing slower-paced games or combining gaming with other tasks such as streaming or content creation. We got our first look at the Odyssey G8 in action with Cyberpunk 2077, and the increased resolution made a significant difference, even if the system struggled to keep up. City streets and distant textures appeared much richer and more detailed.

The Odyssey G8 costs roughly $1600, which places it firmly in the premium bracket. We’re talking about a monitor designed specifically for enthusiasts who already have some serious hardware in their setups and want a display that will evolve with them as their gear improves. The IPS panel sacrifices perfect blacks for OLED, but the trade-off results in improved brightness constancy and crisper text, which is a benefit for everyone who isn’t simply a hardcore gamer.
[Source]

Three months ago, OpenAI cofounder Greg Brockman told me his concerns about a mounting public relations crisis facing artificial intelligence companies: Despite the popularity of tools like ChatGPT, an increasingly large share of the population said they viewed AI negatively. Since then, the backlash has only intensified.

College commencement speakers are now getting booed for talking about AI in optimistic terms. Last month, someone threw a Molotov cocktail at OpenAI CEO Sam Altman’s San Francisco home and wrote a manifesto advocating for crimes against AI executives. No one has more to lose from this reputation crisis than OpenAI.

The person tasked with trying to fix it is Chris Lehane, OpenAI’s chief of global affairs and a veteran political operative. I sat down with him this week to discuss what I’d argue are his two biggest challenges yet: convincing the world to embrace OpenAI’s technology, while at the same time persuading lawmakers to adopt regulations that won’t hamper the company’s growth. Lehane views these goals as one in the same.

“When I was in the White House, we always used to talk about how good policy equals good politics,” says Lehane. “You have to think about both of these things moving in concert.”

After working on crisis communications in Bill Clinton’s White House, Lehane gave himself the nickname “master of disaster.” He later helped Airbnb fend off regulators in cities that viewed short-term home rentals as existing in a legal gray area, or as he puts it, “ahead of the law.” Lehane also played an instrumental role in the formation of Fairshake, a powerful crypto industry super PAC that worked to legitimize digital currencies in Washington. Since joining OpenAI in 2024, he’s quickly become one of the company’s most influential executives and now oversees its communications and policy teams.

Lehane tells me public narratives about how AI will change society are often “artificially binary.” On one side is the “Bob Ross view of the world” that predicts a future where nobody has to work anymore and everyone lives in “beachside homes painting in watercolors all day.” On the other is a dystopian future in which AI has become so powerful that only a small group of elites have the ability to control it. Neither scenario, in Lehane’s opinion, is very realistic.

OpenAI is guilty of promoting this kind of polarizing speech in the past. CEO Sam Altman warned last year that “whole classes of jobs” will go away when the singularity arrives. More recently he has softened his tone, declaring that “jobs doomerism is likely long-term wrong.”

Lehane wants OpenAI to start conveying a more “calibrated” message about the promises of AI that avoids either of these extremes. He says the company needs to put forward real solutions to the problems people are worried about, such as potential widespread job loss and the negative impacts of chatbots on children. As an example of this work, Lehane pointed to a list of policy proposals that OpenAI recently published, which include creating a four-day work week, expanding access to health care, and passing a tax on AI-powered labor.

“If you’re going to go out and say that there are challenges here, you also then have an obligation—particularly if you’re building this stuff—to actually come up with the ideas to solve those things,” Lehane says.

Some former OpenAI employees, however, have accused the company of downplaying the potential downsides of AI adoption. WIRED previously reported that members of OpenAI’s economic research unit quit after they became concerned that it was morphing into an advocacy arm for the company. The former employees argued that their warnings about AI’s economic impacts may have been inconvenient for OpenAI, but they honestly reflected what the company’s research found.

Packing Punches

With public skepticism toward AI growing, politicians are under pressure to prove to voters they can rein in tech companies. To combat this, the AI industry has stood up a new group of super PACs that are boosting pro-AI political candidates and trying to influence public opinion about the technology. Critics say the move backfired, and some candidates have started campaigning on the fact that AI super PACS are opposing them.

Lehane helped set up one of the biggest pro-AI super PACs, Leading the Future, which launched last summer with more than $100 million in funding commitments from tech industry figures, including Brockman. The group has opposed Alex Bores, the author of New York’s strongest AI safety law who is running for Congress in the state’s 12th district.

All-in-one wireless hi-fi systems are no longer the polite little lifestyle boxes hiding on credenzas and apologizing for themselves. In 2026, they are becoming one of the most important categories in premium audio, with Focal’s Mu-so Hekla, and Ruark’s R810 Radiogram, all pushing the idea that better sound no longer has to arrive with a rack of separates, a snake pit of cables, and a marriage counselor/interior decorator on standby.

DALI is now stepping into that fight with VEGA, a $4,500 single box wireless sound system built for listeners who want real hi-fi performance without turning their living room into a shrine to black aluminum. At this price, VEGA will raise some eyebrows, probably in a very restrained Scandinavian manner, but that is where the market is headed: fewer boxes, better industrial design, and sound quality that has to justify more than just convenience.

“As listening habits evolve, more people are enjoying music than ever before thanks to unlimited access to high-quality audio,” said Krestian Pedersen, DALI Head of Product Management. “Music is becoming an integral part of people’s lives, and the key to it all is convenience. It has to be easy, but without compromising on quality. Our goal with VEGA was to create a product that fits the way people live and access music in their daily lives. We wanted to make a product that people want to keep turned on all the time.”

DALI also has an advantage here that many lifestyle audio brands do not. The Danish manufacturer builds almost everything in-house, including drivers, cabinets, and even the small hardware that holds the system together. That level of control has helped DALI develop loudspeakers that blend clean Scandinavian industrial design with real engineering substance.

Over the years, DALI speakers have consistently earned Editors’ Choice recognition because they deliver on the fundamentals: driver quality, cabinet execution, coherent voicing, and strong sonic performance for the money. VEGA is aimed at the same buyer, but with a different brief: deliver a high-quality DALI listening experience from one box that people can leave on, live with, and actually want in the room.

VEGA keeps the Scandinavian design brief intact with real wood veneer finishes, anodized aluminum details, and custom woven fabric. It will be available in Dark Oak and Natural Oak, which should help it blend into actual living spaces rather than looking like a lab instrument that wandered into the dining room.

DALI has also paid attention to the physical controls. The volume wheel uses glass, acrylic, and anodized aluminum, along with an aerospace-grade ball-bearing mechanism designed to give it a smoother, more precise feel.

Uses Ten In-House Drivers and Adaptive Stereo Processing

Every major part of VEGA has been developed by DALI’s engineering team, including the drive units, amplification, and DSP platform. The system uses ten in-house-developed drivers, including ultra-light 25mm soft-dome tweeters with low-viscosity ferrofluid and a large rear chamber designed to reduce resonant frequencies.

DALI has also arranged the bass/midrange drivers in a back-to-back configuration to help reduce cabinet resonance. That matters in a single-box wireless system, where the cabinet has to do a lot of work without becoming part of the performance in the wrong way.

Power comes from 400 watts of amplification, delivered through eight 50-watt channels. VEGA also uses paper-and-wood-fibre cones, low-loss surrounds, and passive bass radiators to support low-frequency output and overall balance. DALI claims VEGA delivers best-in-class bass performance, which is exactly the kind of claim that deserves a listening session before anyone starts slow-clapping in a black turtleneck at a $1 million speaker launch in Aalborg.

The other key technology is DALI Adaptive Stereo Enhancement, or ASE, a proprietary in-house-developed system that is currently patent pending. ASE is designed to create a wider stereo presentation from a single speaker by adapting in real time to the incoming signal.

DALI VEGA Adjusts for Placement and Orientation

VEGA can be used freestanding or wall-mounted in portrait or landscape orientation. DALI’s Adaptive Orientation Adjustment, or AOA, automatically adjusts the speaker’s output based on how it is positioned, including stereo mapping and spatial presentation.

Users can also adjust placement settings based on proximity to walls or corners, which should help VEGA perform more consistently in real rooms. That matters at $4,500, because “just put it anywhere” is usually where good sound goes to die behind a ceramic vase.

BluOS, HDMI, AirPlay 2, and TIDAL Connect

VEGA is built around BluOS, giving it high-resolution wireless streaming, multiroom playback, internet radio, and app control across compatible BluOS-enabled products.

DALI also gives VEGA a useful mix of inputs, including HDMI, analog, optical digital, USB audio, and Bluetooth, which makes it more flexible than a wireless speaker that only wants to live inside an app. It can connect to TVs, digital sources, and analog components, although turntable users will still need to confirm whether their deck has a built-in phono stage or use an external one.

Streaming support includes Spotify Connect, TIDAL Connect, and Apple AirPlay 2, with preset buttons on the unit for quick access to favorite sources or playlists. The OLED display also rotates with the speaker’s orientation, which is a small but smart touch. At $4,500, “small but smart” should be part of the admission price.

DALI VEGA Key Specifications

System type:
Single-box wireless hi-fi system with active 2-way crossover

Drivers:
4 x 25mm soft-dome tweeters
4 x 4.5-inch paper-and-wood-fibre bass/midrange drivers
2 x rectangular 3 x 6-inch passive radiators

Amplification:
400 watts total
8 x 50-watt Class D BTL amplifier channels

Frequency response:
32Hz to 22.7kHz, ±3dB

Maximum SPL:
110dB at 1 meter

Bass tuning:
Passive radiator tuning frequency: 32Hz
Subwoofer output with 120Hz low-pass filter

Streaming and wireless:
BluOS
Spotify Connect
TIDAL Connect
Apple AirPlay 2
Bluetooth with AAC, aptX, and aptX HD

Supported audio:
16-bit to 24-bit audio
32kHz to 192kHz sample rates
BluOS supports FLAC, ALAC, WAV, AIFF, MP3, AAC, OGG, WMA, MQA, and DSD256 with DSP-to-PCM conversion

Inputs:
HDMI ARC
Stereo RCA analog input
Optical digital input
USB Audio/Service for HDD or USB drive
Bluetooth
BluOS

Outputs:
Subwoofer output
USB power, 5V/1A

Placement options:
Freestanding in free space
Freestanding near a wall
Freestanding close to a corner or wall
Wall-mounted in landscape or portrait orientation

Key features:
Adaptive Orientation Adjustment
Adaptive Stereo Enhancement
Up to 40 programmable presets
Direct or Custom EQ
Rear-wall distance adjustment
HDMI audio delay
Input sensing
OLED display that rotates with orientation

Dimensions:
5.63 x 26.90 x 9.57 inches
143 x 683 x 243 mm

Weight:
19.18 pounds
8.7 kg

Finishes:
Dark Oak
Natural Oak

The Bottom Line

DALI VEGA is not just another wireless speaker with better clothes. It is a $4,500 all-in-one wireless hi-fi system built around ten in-house-developed drivers, 400 watts of amplification, BluOS streaming, HDMI ARC, AirPlay 2, TIDAL Connect, Spotify Connect, and DALI’s Adaptive Stereo Enhancement and Adaptive Orientation Adjustment technologies.

What makes it interesting is DALI’s control over the hardware. The company builds its own drivers, cabinets, and key components, which gives VEGA a stronger engineering foundation than many lifestyle-first wireless systems. The real trick is whether DALI can deliver a convincing stereo presentation and proper low-frequency performance from one box without making it sound like digital wizardry wearing Danish furniture.

What is missing? There is no HDMI eARC, no Wi-Fi 7 future-proofing, no listed Dolby Atmos support, and turntable users will still need a phono preamp unless their deck has one built in. At $4,500, those omissions matter.

VEGA is for design-conscious music listeners who want serious hi-fi performance without separates, speaker cables, or a rack full of gear. It will make its public debut in Vienna next month, but it does not go on sale until September, which gives everyone enough time to decide whether one box can replace a system, or just rotate beautifully while trying.

Related Reading:

Following the initial tease of its Thus AI chip, Anker announced the first earbuds with the component during its Anker Day event. The Soundcore Liberty 5 Pro and Liberty 5 Pro Max both utilize the chip for what the company calls Whisper Clear voice performance. Of course, the usual earbud features are here — active noise cancellation (ANC), voice controls and personalized sound — but there are some unique tools available on these models as well. Those include an AI Note-Taker on the Liberty 5 Pro Max and touchscreen charging cases for both. 

Let’s start with what the Thus chip does on both sets of Liberty 5 Pro earbuds. The silicon is used alongside eight microphones and two bone conduction sensors to ensure optimal voice performance in noisy locations. By also employing an AI model, the system separates the speaker’s voice from background noise so they’re heard clearly on the other end of calls. Anker says the bone conduction sensors allow the setup to capture the user’s voice a lower volumes, so long as they’re in “moderately quieter settings.” The Thus chip also powers voice commands, offering 20 options for various controls and tasks, with Anker claiming a response time of under a second. 

Both the Liberty 5 Pro and Liberty 5 Pro Max have Soundcore’s Adaptive ANC 4.0 that uses those eight mics to process audio data at up to 384,000 times per second. This action continuously monitors both external noise and any sounds that may enter the ear canal before adjusting the ANC in real time. Anker says its noise-canceling algorithm can combat a broad range of noises — including human voices. What’s more, the company claims the ANC performance on its new earbuds is up to 2x more effective than the Liberty 4 Pro. 

Anker

To customize the sound to your ears, both sets of earbuds offer HearID 5.0. This feature creates a personalized EQ based on a brief hearing test. With the help of AI Sound Enhancement, the earbuds provide audio restoration that can recover up to 65 percent of “the quality” that’s typically lost to Bluetooth compression.

In terms of battery life, both the Liberty 5 Pro and Pro Max will last up to 6.5 hours on a charge with ANC enabled. When you factor in the charging cases, the total use time is 28 hours for each model. Multipoint Bluetooth connectivity is also onboard both, as is Bluetooth 6.1, Apple Find My and Google Fast Pair compatibility. Plus, both sets of earbuds are IP55 rated for dust and water resistance. 

There is one main area where the two new models differ is their smart cases. The Liberty 5 Pro case has a 0.96-inch TFT touchscreen on its front edge while the Liberty 5 Pro Max case has a much larger 1.78-inch AMOLED display. The Pro Max is also equipped with an AI Note-Taker that’s capable of generating transcripts, including identifying speakers and action items, after meetings in the Soundcore app. This feature isn’t available on the Liberty 5 Pro, but Anker does offer AI-based translation on both models. Translation is available on the earbuds with the Liberty 5 Pro, and it’s available on both the earbuds and the case with the Liberty 5 Pro Max. 

Both the Liberty 5 Pro and Liberty 5 Pro Max are available today for $170 and $230 respectively. The Liberty 5 Pro is available in black, blue, pink and white while the Pro Max will come in either black or gold.  

ClickUp, the $4 billion productivity platform, has cut 22 per cent of its workforce. CEO Zeb Evans announced the layoffs in a post on X, framing them not as a cost-cutting exercise but as a structural bet on AI. The savings, he said, will flow back to the employees who stay in the form of million-dollar salary bands.

Evans called the new structure a “100x org.” The premise is that AI agents have changed what it takes to build software, and the roles required to operate at the highest level are now fundamentally different. Incremental improvements to existing systems will not get ClickUp there, he argued. The company needs to rebuild rather than iterate.

The restructuring follows months of aggressive AI adoption inside ClickUp. A Fortune profile published days before the layoffs revealed that the company now runs roughly 3,000 internal AI agents across its departments, a 3:1 ratio of agents to employees. Evans had already mandated that staff go through an AI agent trained to stand in his place before contacting him directly.

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!

Advertisement

Evans outlined three categories of employee he sees as essential to the new model. The first is “builders,” which he splits into 10x engineers and 10x product managers. His claim is blunt: the best engineers are not writing code any more. They are directing agents that write code. The skill that matters is judgment, the ability to orchestrate and review. AI makes the best engineers wildly more productive, he wrote, while everyone else using AI slows them down.

He called this the “great reckoning of AI coding” and said every company will face it soon. Companies celebrating 500 per cent more pull requests are generating volume, not outcomes. More code, in his view, is just another bottleneck.

The second category is “system managers,” or agent managers. These are people who automate their own jobs with AI and then become owners of the systems they built. Evans argued that anyone who automates their role will always have a job. The underlying systems, not the individual tasks, are what matter.

The third is “front-liners,” the people who spend their time with customers. In a world saturated with AI communication, Evans said, human contact becomes the one bottleneck companies should not try to replace. Front-liners should spend nearly 100 per cent of their time in meetings with customers, while the systems around those meetings are fully automated.

Product management and design, he added, are merging. Designers with customer focus become more like product managers. Product managers with UX intuition become more like designers. The bottleneck of user research is gone, he claimed, because a single mention to an agent can kick off and analyse a research cycle.

The most provocative element is the compensation model. ClickUp is introducing salary bands that reach $1 million per year in cash. The path is available to nearly anyone in the company who produces “100x impact” by creating or managing AI systems. In a world where the best people create 100 times more output, Evans argued, companies cannot afford to lose them and should aim to retain them for decades.

The announcement lands in the middle of a brutal stretch for tech workers. The industry has shed more than 100,000 jobs across roughly 250 events in 2026 so far. Meta cut 8,000 roles the same week despite record revenue. Oracle eliminated up to 30,000 to fund AI infrastructure. GitLab restructured for the “agentic era.” The pattern is consistent: companies report record performance and cut headcount simultaneously, redirecting savings into AI.

Evans’s framing is more explicit than most. Where other CEOs use euphemisms about efficiency and realignment, he is making a direct argument that the roles being eliminated are structurally obsolete. Whether that is candour or hubris will depend on whether the 100x org delivers the outcomes he is promising.

ClickUp reported roughly $300 million in annual recurring revenue as of 2025 and has been eyeing an IPO. The company acquired AI coding platform Codegen late last year. With AI reshaping the economics of developer tools and productivity software, Evans is betting that a smaller, better-paid workforce directing thousands of agents will outperform the company it replaces.

Not everyone is convinced. In China, courts have ruled that replacing workers with AI is not legal grounds for dismissal. In the US, no such protection exists. For the 22 per cent of ClickUp employees who lost their jobs this week, the distinction matters.