F5 has released security updates to address a critical BIG-IP APM zero-day vulnerability being exploited in remote code execution attacks.

BIG-IP APM (short for Access Policy Manager) is the company’s centralized access management proxy solution that helps admins secure access to their organizations’ networks, applications, cloud, and application programming interfaces (APIs).

Tracked as CVE-2026-94127, the flaw affects instances configured as an OAuth Authorization Server when a BIG-IP APM access policy and an OAuth profile are configured on a virtual server.

“We have learned that this vulnerability has been exploited,” F5 warned in a security advisory published on Tuesday. “Deployments using APM strictly as an OAuth Client / Resource Server (without OAuth authorization server profiles configured) are not affected by this vulnerability.”

The company advised customers to review systems for indicators of compromise if they detect a combination of multiple OAuth authentication failures and suspicious commands, shortly followed by a TMM SIGABRT.

F5 also shared mitigation measures for admins who can’t immediately install the security updates, which require applying an iRule (available from F5 Support) to the affected BIG-IP APM virtual server.

Internet threat monitoring non-profit Shadowserver currently tracks over 14,700 IP addresses with BIG-IP APM fingerprints. However, there is no information on how many have already been patched or are honeypots.

On Tuesday, the Cybersecurity and Infrastructure Security Agency (CISA) also added CVE-2026-94127 to its Known Exploited Vulnerabilities (KEV) Catalog and ordered U.S. federal agencies to secure their networks against this flaw by Friday.

“These types of vulnerabilities are a frequent attack vector for malicious cyber actors and pose significant risks to the federal enterprise,” the cybersecurity agency warned.

Cybercrime and state-backed threat groups have often exploited F5 vulnerabilities in recent years. For instance, attackers have targeted security flaws in F5 products to breach corporate networks, hijack devices, ​​​​​​map internal servers, deploy data-wiping malware, and steal sensitive documents.

F5 also disclosed in October 2025 that state-sponsored hackers breached its systems in August 2025 and stole undisclosed BIG-IP security source code and vulnerabilities.

Since November 2021, CISA has flagged eight actively exploited F5 vulnerabilities, four of which have also been abused in ransomware attacks.

F5 is a Fortune 500 company that provides cybersecurity, application delivery networking (ADN), and other services to more than 23,000 customers worldwide, including 48 of the Fortune 50 companies and 80% of the Fortune Global 500.

Join Mikko Hyppönen and security leaders from the NFL, CHANEL, and Atlassian for a two-hour digital summit on what AI-speed attacks change, what defenders should stop doing, and how to validate, decide, fix, and re-validate at machine speed.

Save your seat

From the day I started Heidi, the ambition was always bigger than writing doctor’s notes.

Dr Thomas Kelly said that on Tuesday, as the company he cofounded announced US$340m of new funding.

The money arrives in two parts. In its announcement, Heidi said Blackbird led a US$100m Series C that values the Melbourne company at US$900m. Phoenix Court, Point72 Private Investments and Headline also took part.

That price is almost double the US$465m Heidi carried last October, when Point72 led a US$65m Series B.

The other US$240m is not equity

The larger half comes from General Catalyst’s Customer Value Fund, and it buys no shares.

Olivia Poh reported for Bloomberg that the CVF is a debt-like instrument. General Catalyst holds no stock and no warrants in Heidi. It funds sales and marketing, then takes a capped share of the revenue that spending produces, on top of getting its money back.

Pranav Singhvi, who cofounded the fund, said it exists so companies with good unit economics can push their growth engine without spending their own balance sheet on it.

Heidi has now raised US$436.6m in total.

The numbers behind the round

Annual recurring revenue went from US$1m to US$50m in two years, reaching that mark in April.

Heidi says its product has supported more than 175 million patient visits, up from 73 million at the Series B, and more than 67 million clinical hours. It puts weekly volume at about 2.8 million patient interactions, across 110 languages and 190 countries. A year ago it counted 116 countries.

At the Series B last October, Heidi reported about two million patient interactions a week across 116 countries. Those are the figures the new ones are measured against.

Enterprise activation sits at around 62%, by the company’s own measure. Heidi employs close to 600 people.

Kelly founded the company in Melbourne with Waleed Mussa and Yu Liu. Blackbird has held a stake since the pre-seed in 2021 and leads this round.

Michael Tolo of Blackbird, which has backed the company since its pre-seed in 2021, said generic AI will not close the healthcare deficit because clinicians ignore it.

Kelly told Bloomberg the company plans to add as many as 150 staff over the next year, and to push into Asia and the Middle East.

Where Heidi already runs

In the UK, Heidi is sole supplier to NHS England Midlands, in what the company calls the largest clinical AI procurement in NHS history. In the US it is used by Beth Israel Lahey Health in Massachusetts. In New Zealand it runs in every emergency department in the country.

Australian deployments include the Royal Children’s Hospital Melbourne, Children’s Health Queensland and Metro South Health.

Britain is buying clinical AI at scale at the moment. Palantir’s £330m NHS contract is heading for renewal with a former deputy Labour leader newly on its payroll.

What the money pays for

Kelly said the round funds a move from documentation to agents that take on clinical work under supervision. He told Bloomberg that within six to twelve months Heidi should handle most of what a doctor asks of it.

Heidi has built its own models rather than renting frontier ones. It says transcription and note generation, which account for most of its compute, now run almost entirely on in-house technology.

Beyond the scribe it has shipped Evidence, which answers clinical guideline questions and has handled more than 10 million queries since March. It has also shipped Remote, a wearable for audio capture, and Dictate.

Part of the round pays for clinical evidence, quality management systems and regulatory submissions. Heidi says it will work with regulators on the appropriate pathway.

The General Catalyst money is tied to go-to-market spending by the structure of the fund, so the equity is what pays for the rest.

Not in the UK or EU

Heidi added one line under its investor quotes. The coming product capabilities will not be available in the UK and EU. The company gave no reason and no timeline.

It does list the paperwork it holds: ISO 27001, SOC 2 Type II, Cyber Essentials Plus and ISO 42001, plus alignment with NHS requirements, GDPR, HIPAA and the Australian Privacy Principles.

Europe has a route for autonomous clinical AI, and it is slow. Vara spent years securing CE certification for AI that reads breast screening images without a second radiologist.

The scrutiny around clinical AI

We reported in August that NHS AI scribes were making errors in patient records, with Healthwatch raising concerns and the MHRA involved. In a separate case, an AI receptionist mishandled a stroke patient.

Heidi says every agentic decision stays under clinician control. Its clinical research team runs testing, validation and published research ahead of release.

Bloomberg points to a 2023 Pew Research Center survey in which 60% of American adults said they would be uncomfortable if their doctor relied on AI to diagnose them.

Other scribe rounds this year

Tandem Health raised $100m on 15 September to turn its own AI scribe into clinic software. That is the same amount of equity, for the same strategic move, one week earlier.

Bloomberg notes two more. Forus took $150m this month at a $3bn valuation, and Latent raised $80m earlier in the year.

Kelly’s pitch against general-purpose assistants is retention. He told Bloomberg that healthcare involves a lot of setup and configuration, and that the company’s data shows customers stay.

The World Health Organization projects a shortfall of 11 million health workers by 2030. Heidi cites that gap as the market it sells into.

Shadow IT includes hardware, software, and services running outside IT and security teams’ visibility or approval.  It can include unapproved applications employees install, browser extensions with excessive permissions, and endpoints provisioned once and never enrolled in any monitoring platform. These unmanaged assets create visibility gaps because existing security controls do not monitor them.

Many organizations rely on network discovery scans to measure asset coverage. While useful, network discovery only identifies endpoints that respond during the scan window. Powered-off devices, isolated network segments, and software that does not expose listening ports can remain invisible. As a result, network discovery measures network reachability, not monitoring coverage.

Wazuh is a free and open source security platform that unifies SIEM and XDR capabilities across endpoints and cloud workloads. It collects system inventory data directly from each monitored endpoint, allowing security teams to compare what network scans report against what the Wazuh-monitored endpoints report.

This comparison helps identify unmanaged endpoints, unauthorized software, and monitoring gaps.

Why shadow IT resists discovery

Shadow IT persists because the tools that report findings often can’t observe the assets in question. Understanding where the gaps form helps security teams decide which controls to extend.

• Unmanaged endpoints: A workstation that is reimaged and never re-enrolled, or a virtual machine built for a short-lived project and left running, produces no telemetry. It does not appear in patch reports, vulnerability findings, or alert data, because no agent is installed to monitor them.
• Unapproved applications on managed endpoints: Remote access tools, file sharing clients, and productivity utilities installed outside the approved software baseline introduce risk on endpoints that are otherwise fully monitored. The endpoints are visible, but the software running on them is not always reviewed against policy.
• Software that opens no network port: Unauthenticated network discovery can infer services from the ports a host exposes, but it cannot reliably identify software that does not listen for inbound connections. Browser extensions, locally installed utilities, and remote access tools that initiate outbound connections to a broker are not observable through that method.
• Devices that cannot run an agent: Printers, switches, IP cameras, and other network devices require inventory and monitoring methods other than an endpoint agent.

Addressing these categories requires endpoint telemetry and additional data sources for devices that cannot run an agent.

Closing the blind spot with Wazuh

Wazuh helps organizations reduce shadow IT exposure through continuous system inventory collection, centralized analysis, and correlation of inventory data with vulnerability and policy information. The following capabilities help security teams identify unmanaged assets and unauthorized software across their environment.

• Continuous system inventory: The Wazuh agent collects hardware details, operating system information, installed packages, network interfaces, listening ports, running processes, services, users, groups, and browser extensions from each monitored endpoint. The Wazuh agent on Windows endpoints additionally reports installed updates. The Wazuh agent forwards this data to the Wazuh server, which processes it and stores it in the Wazuh indexer as the endpoint’s current state. The system inventory capability is enabled by default, runs an initial scan when the Wazuh agent starts, and rescans on a configurable interval that defaults to one hour.
  

  

• Centralized visibility across the environment: Inventory data from every monitored endpoint is aggregated into dedicated indices and presented on the Wazuh dashboard under IT Hygiene, organized into Dashboard, System, Software, Processes, Network, Identity, and Services sections. Security teams can query the entire data from a single interface rather than inspecting endpoints individually. The Network section includes a Traffic tab listing the ports actively listening on each endpoint. This is a host-side view of the same information a network scan attempts to infer remotely.

• Visibility into browser extensions and endpoint services: Wazuh 4.14.0 and later collect inventory data about browser extensions, endpoint services, users, and groups. The browser extension model is unified across Windows, macOS, and Linux, allowing security teams to identify extensions with broad permissions across the estate from a single query. The services inventory normalizes Windows services alongside Linux systemd units, which helps teams find unapproved software configured to start automatically.

• Coverage for devices that cannot run a Wazuh agent: Wazuh provides agentless monitoring over SSH for routers, firewalls, switches, and Linux or BSD systems. For devices that support neither an agent nor an SSH session, the Wazuh server accepts syslog directly from network appliances and other equipment. These capabilities provide additional visibility into devices that cannot run a Wazuh agent. However, SSH and syslog monitoring provide only the information exposed by the configured commands, files, and device logs.
• Monitoring agent coverage itself: Agent enrollment status measures coverage among endpoints already registered with the Wazuh server. It does not identify unknown endpoints that were never installed or enrolled. A Wazuh agent reports one of four states: never connected, pending, active, or disconnected. Agents that enrolled but never connected indicate deployments that did not complete. An agent moves to disconnected when no keep-alive message arrives within the configured window, which defaults to 15 minutes. Reviewing these states helps security teams identify endpoints assumed to be monitored but not reporting.
• Correlating inventory against vulnerabilities and policy: Inventory data supports security decisions when compared against a reference. Wazuh vulnerability detection correlates collected software inventory with vulnerability content from the Wazuh Cyber Threat Intelligence (CTI) platform. It does not probe endpoints or sweep the network, so the analysis runs centrally on data the endpoints have already reported. The same inventory can be joined against external data sources to identify end-of-life software that no longer receives security updates. 

• Threat hunting across inventory data: Because inventory is stored centrally and indexed, security teams can query it during investigations rather than collecting data on demand. The Wazuh blog post Threat hunting using inventory data collected by Wazuh demonstrates this with custom rules, CDB lists of unauthorized software and ports, and queries run against the inventory. This approach surfaces assets that no single alert flags on its own.
• Detecting and removing unauthorized applications: Security teams can configure custom Wazuh rules to detect selected unauthorized applications and use a tested Active Response script to remove them from monitored endpoints. The Wazuh blog post Managing shadow IT with Wazuh covers the full configuration, including the custom detection rules and the active response script that uninstalls unapproved applications from monitored endpoints.

Conclusion

Shadow IT is a visibility problem before it is a policy problem. Organizations cannot enforce a software baseline on assets they cannot see, and network scanning alone shows only the systems that respond during a scan.

Wazuh helps reduce this visibility gap by collecting inventory from monitored endpoints and ingesting data from configured agentless sources. Endpoint inventory shows what is installed, running, and listening on each host, while agentless monitoring and syslog ingestion extend visibility to supported devices that cannot run a Wazuh agent. Wazuh agent enrollment states also help security teams measure monitoring coverage instead of assuming it. 

With this data indexed centrally, security teams can identify vulnerabilities, investigate inventory across monitored endpoints, compare software with lifecycle information, and initiate configured remediation for applications outside the approved software baseline.

Discover more about Wazuh by exploring its documentation and joining its growing community of professionals.

Sponsored and written by Wazuh.

A three-and-a-half-minute video released this weekend made an N64 first-person shooter look like it belonged on the Super Nintendo, a 16-bit machine that many grew up with. Bond stands on the Arkhangelsk dam, the gun is in the lower corner of the frame, and the concrete bunkers pass by in chunky polygons while fog eats the far cliffs.

Retroports released a new video on their channel on September 20th. The video is simply titled ‘GoldenEye SNES Test’, and a small but devoted audience of a few hundred subscribers watches to see what they’re up to. The video in question is a live capture from a Mesen emulator, showing the opening and Dam missions from a 2MB Super FX2 cartridge running in polygon rendering mode, with some interesting features like fog and occlusion culling in action. Unfortunately, with such a restricted texture budget, just the first part of the level is rendered. What you may not know is that Claude Opus did the most of the work on the renderer, while ChatGPT Astra addressed a few other bugs.

For those who can remember back to the 1990s, Super FX2 was the newest generation of Argonaut and Nintendo’s collaboration, a processor that was shoehorned onto a few late SNES carts. Star Fox was the first game to make use of the original Super FX processor, which operated at around 10.7 MHz. On the other side, titles like Yoshi’s Island, DOOM, and Stunt Race FX required the extra horsepower of the Super FX2, which ran at 21.47 MHz, to power through all of the 3D arithmetic that the 65816 CPU alone couldn’t manage. Retroports has been playing with a 2MB Super FX2 cart inside Mesen, and they’ve used the extra power to test out a first-person camera, textured walls, and an animated gun sprite.

The Dam is the only mission that survives the cut, which is logical given how simple it is to texture a large concrete wall in comparison to all the offices and city blocks on the other levels. Then there’s occlusion culling, which helps by hiding geometry that the camera can’t see, and fog, which helps by reducing the draw distance so the chip doesn’t have to paint the entire valley every frame. Still, you can see the 1997 intro to Rare’s GoldenEye, even if it’s running on a device that’s a few generations behind the Nintendo 64.

Following Donkey Kong Country, Nintendo Power reported that Rare was working on GoldenEye for the Super NES, with initial plans to adopt a similar pre-rendered graphic style. However, the team eventually switched to a 64-bit processor, which resulted in the first-person shooter that redefined multiplayer in living rooms for a decade. Anyway, Retroports has managed to recreate some of the latter design on the earlier machine, utilizing the same chip that Nintendo gave for Star Fox plus a handful of language models to improve rendering.

To be totally honest, no one has been able to obtain a ROM for this game because, as far as we know, it has only ever been loaded onto the original hardware by the Retroports team themselves. Until we have a file that other people can run, this clip will have to suffice as proof in the pudding. Retroports describes it as a test to determine if 3D on the SNES may have been a more economically viable idea.
[Source]

Smart glasses may look cool but there’s absolutely no place in our schools or classrooms for them.

Sabine Winton said that on Tuesday. Western Australia’s education minister was announcing what she called the country’s most comprehensive ban on the devices.

It is not a policy statement. The same day, the department’s director general Jay Peckitt issued an instruction to every public school principal in the state. It takes effect on 12 October, the first day of Term 4.

What the instruction actually orders

The CEO Instruction is short and unusually specific.

Principals must stop students and staff wearing, using, charging or storing the glasses on school premises. The same applies at any school organised event or activity. The charging clause matters, because it closes the locker.

Principals must also bar parents, carers, visitors and contractors from bringing the devices onto school grounds at all. The instruction says the ban should reach government residential colleges too, through separate mechanisms.

It runs on real legal authority rather than guidance. The director general issues it under section 233 of the School Education Act 1999. Principals enforce it under regulations 69 and 70. Those let a principal direct people on school premises, and bar an item from coming in. Both rules say the school must display the order publicly.

The risk nobody expected on the list

The instruction names six risks. Four are the ones you would guess: privacy, child safety and wellbeing, unauthorised recording, and cyber security.

The fifth is disruption to teaching. The sixth is examination and assessment integrity.

That last one reframes the device. The instruction’s own definition runs well past a camera. It lists audio recording, video, photography, artificial intelligence functionality and live streaming. Then internet connectivity, facial recognition, information retrieval and communication.

Principals can still approve the glasses for medical, accessibility or assistive technology use, and for educational programmes they authorise themselves. Students who break the rule fall under the student behaviour policy, staff under staff conduct and discipline. Principals can tell visitors and contractors to take the devices off, and to leave if they refuse.

No incidents, on the record

Winton said the education department has recorded no instances of smart glasses being used irresponsibly in WA schools.

Prevention is the stated driver. Callum Liddelow and Courtney Withers reported the ban for ABC News. Winton pointed to growing community concern about careless, criminal and malicious use, not to anything that has happened in a classroom.

Kya Graves runs the WA Principals Federation. She told the ABC that principals had been writing their own school-level policies, with no departmental decree to work from. Some had incidents on their sites and put measures in place themselves.

Tama Leaver teaches internet studies at Curtin University. The worst outcome would be indecent recording of children, he told the ABC. He raised deepfakes as the second risk, and said it takes about 30 seconds of footage to make one with rudimentary tools.

Canberra asked the councils to do it

The federal government is the one part of this not acting.

Attorney-General Michelle Rowland ruled out an import ban on 31 August. Erin Handley reported for the ABC that Rowland said the government was not consulting on one, though it was not ruling anything out on other measures.

In its place she offered a privacy overhaul. The draft laws carry a right to erasure from large platforms and search engines, a fair and reasonable test on data collection, and meaningful consent rules. An identity document service called IDLock is due in 2027.

Rowland also encouraged councils to restrict the glasses in their own facilities, and suggested retailers could police change rooms.

The eSafety commissioner put the other side. The devices help people with disability, through speech to text for deaf users and scene descriptions for people with low vision.

The councils took the invitation

Matthew Taylor reported for Lawyers Weekly that the City of Sydney and Canterbury-Bankstown have now barred the glasses from their pools and fitness facilities, joining Brisbane and Yarra.

Councillor Matthew Thompson put the motion on Monday. It directs the chief executive to review the council’s conditions of entry and update them, covering leisure centres, libraries, council childcare centres and community centres. The document notes the pools already excluded the glasses.

Canterbury-Bankstown got there first and wrote in enforcement. Its staff can order patrons to remove the eyewear, confiscate it, and move on anyone who refuses. Women and parents raised the concerns that started it.

Brisbane barred non-consensual use of camera-enabled devices at its 21 public pools, a category taking in phones and ordinary cameras. Gold Coast is considering the Brisbane model, Adelaide is weighing its own, and the City of Vincent in Perth ordered a report last week.

Thompson wants the import ban Rowland declined. He told TechRadar that Canberra has left the problem with councils. Federal crossbenchers asked for the same ban in late August.

The office is next

Employment lawyers are now telling Australian employers to write their own rules.

Paul O’Halloran, a partner at Dentons, told Lawyers Weekly the technology exists in public and workplace settings for one practical reason, which is to record people without their knowledge or consent. He said it should be banned outright.

He called the technology insidious, because unlike a phone or a camera there is no obvious moment of recording. That is the feature rather than the flaw, he said.

O’Halloran argues employers already carry the risk. Privacy law and work health and safety codes oblige them to protect staff from harm, including covert surveillance. An employee could be recorded all day, in meetings, in break rooms, even in bathrooms and change facilities, and neither the employer nor the person filmed would know.

Michael Byrnes, a workplace relations partner at Swaab, put it more plainly. Employers should have no hesitation in banning the glasses, he said, and telling staff not to wear them is a reasonable and lawful direction.

His specific worry is audio. Secretly recording private conversations potentially breaches surveillance devices legislation.

Byrnes allows two exceptions. Employees with a visual impairment should be able to use them as far as the work requires. Where the glasses reduce a genuine safety hazard, that may also justify them, with rules on retaining and using any recording.

Three statutes already cover this ground: the NSW Workplace Surveillance Act 2005, the Privacy Act 1988 and the Fair Work Act 2009.

What the evidence says

A University of Sydney study sampled 350 public Instagram videos shot on smart glasses. It found that around 60% could potentially count as harassment of a person or people. We covered that research in August, and the way the algorithm pushed those videos.

Independent testing published this week found a second problem. A hacker can get into some ultra-cheap smart glasses without a password.

People have started building their own defences. A Polish developer released an app that detects the glasses over Bluetooth.

Not only Australia

Courts in England and Wales barred the glasses in August. New York state courts did the same in July. In the United States, cinemas, schools and Immigration and Customs Enforcement have all restricted them.

Australia already banned social media for under-16s, and Victoria bars wearables in schools from January 2027. South Australia’s attorney-general, Kyam Maher, is taking advice on whether state law is adequate, and has been talking to Rowland about nationally consistent rules.

The rapid growth of extra-large TVs measuring 85 inches and larger has put real pressure on the home projector category. Today’s biggest TVs are brighter, smarter, and increasingly affordable, making them a compelling alternative for buyers who once would have automatically considered projection.

Projectors still hold one major advantage, however: sheer image size. In dedicated home theaters and larger multipurpose rooms, projection can deliver a 100-inch, 120-inch, or even larger image without pushing prices into the stratosphere.

Following the launch of its higher-end Pro Cinema Q-Series 4K Laser Projectors in 2024, Epson is now bringing more attention to its step-down Home Cinema lineup with the new LS5000, LS6000, and LS7500 4K HDR laser projectors. The three models sit below the flagship Q-Series but retain a number of the technologies and features that made those projectors noteworthy.

The new LS Series models are built around Epson’s 3-Chip 3LCD technology and use a laser light engine. For owners of older lamp-based projectors, the move to laser brings a longer-lasting light source, more consistent brightness over time, and faster startup and shutdown.

The new models complement Epson’s established Home Cinema projector lineup, which is currently headed by the Home Cinema LS11000, introduced in 2022. Epson says the new Home Cinema projectors are designed to balance image performance with flexible installation and easier integration across a wide range of home entertainment spaces.

“As TVs get bigger, they can also become a design challenge, leaving a large black rectangle that dominates the room when not in use. Projection offers a discreet alternative that virtually disappears when the entertainment ends, while creating immersive experiences not possible with a traditional TV – from everyday 16:9 content to cinematic 21:9 and IMAX-style viewing,” said Fernando Tamashiro, product manager at Epson America, Inc. “By expanding our Home Cinema lineup with these new 4K HDR laser projectors, we’re offering an even broader range of solutions for today’s home entertainment spaces. Combining our unique 3-Chip 3LCD technology with our latest laser light source, Epson projectors are the optimal choice for customers looking to create a big-screen experience tailored to their viewing preferences, space and budget.”

Designed for movies, sports, and immersive gaming, Epson’s new Home Cinema laser projector lineup can display 4K HDR images at screen sizes ranging from 40 to 300 inches.

What the Epson LS5000, LS6000, and LS7500 Have in Common

Image Display Technologies: All three projectors use an advanced Epson Picture Processor, proprietary 4K display technology that produces a full 8.29-million-pixel on-screen image via Quad Pixel Shifting, and a 3-Chip 3LCD projection engine powered by a laser-array light source.

Light Output: The three projectors produce between 2,500 and 3,500 lumens of equal Color Brightness (IDMS Rated) and White Brightness (ISO Rated), depending on the model. Epson says the result is a bright, vivid image with strong color performance and deep black levels.

Advanced HDR Capability: Full 10-bit color processing supports HDR10 and HLG content. Additional HDR processing is provided by Epson’s Automatic HDR Scene Correction.

Multiple Aspect Ratios: From 16:9 HDTV content to cinematic 21:9 presentations, the new Home Cinema LS projectors are designed to support a range of screen formats and viewing environments.

Lenses: The Home Cinema LS5000 and LS6000 incorporate Epson Cinema standard-throw lenses, while the Home Cinema LS7500 features a short-throw lens that allows the projector to be positioned closer to the screen than conventional long-throw models. That makes the LS7500 better suited to larger screen sizes in rooms where projector placement is more limited.

High Black Contrast: Producing convincing black levels while maintaining shadow detail has long been a challenge for projectors. Epson specifies dynamic contrast ratios ranging from 800,000:1 to 1,200,000:1 across the three new LS models, depending on the projector. The goal is greater differentiation between bright and dark portions of the image while preserving detail in darker scenes.

HDMI Connectivity: All three projectors include dual HDMI inputs supporting refresh rates up to 120Hz, with one HDMI port providing eARC for audio passthrough to a compatible AV receiver or soundbar.

Gaming: In addition to supporting refresh rates up to 120Hz, the new LS projectors include Auto Low Latency Mode (ALLM) and input lag rated below 20ms. That combination makes them suitable for gaming with compatible consoles and PCs where faster response times are important.

Flexible Installation: A compact design with quiet, front-exiting airflow provides additional placement flexibility. Manual lens shift of ±21% horizontally and ±59% vertically also makes it easier to position the projector in a variety of home theater and multipurpose spaces.

Bluetooth: Bluetooth connectivity is still not standard across the projector category, but the Home Cinema LS5000, LS6000, and LS7500 include Bluetooth 6.0. This allows audio from HDMI-connected sources to be streamed to compatible wireless headphones or speakers.

Comparison

The Bottom Line 

Epson’s position as the world’s top-selling projector brand did not happen by accident. Its lineup stretches from classrooms and conference rooms to dedicated home theaters, and the new Home Cinema LS5000, LS6000, and LS7500 fill an important spot just below the company’s more expensive Pro Cinema Q-Series.

What makes these three interesting is not simply that they use laser light sources or can throw a 300-inch image. Epson combines its 3-Chip 3LCD architecture with equal rated color and white brightness, output ranging from 2,500 to 3,500 lumens, Quad Pixel Shifting for an 8.29-million-pixel 4K image, HDR processing with tone mapping, and gaming features that include 120Hz support and ALLM. The LS7500 also adds a short-throw lens for installations where putting a conventional projector farther back in the room simply isn’t practical.

That combination gives the LS Series a legitimate point of differentiation from many single-chip DLP competitors, where white-light output and color-light output can differ substantially. It also gives buyers much of the functionality they actually need for movies, sports, and gaming without forcing them into Epson’s considerably more ambitious Pro Cinema Q-Series.

Competition from Sony, Hisense, Leica, JMGO, Optoma, and others is substantial, and some alternatives offer more light output, different installation options, or integrated entertainment features. But for someone trying to build a genuinely large-screen home theater without entering flagship-projector territory, the LS5000, LS6000, and LS7500 make a strong case for staying with traditional projection. A 98-inch TV might be getting easier to justify; a 150-inch one remains something of a financial incident.

Pro Tip: None of the three projectors includes a built-in speaker system or projection screen, so those need to be factored into the total system cost. Bluetooth is available for wireless audio, but anyone building a serious home theater will still want to budget for an external audio system and an appropriate screen.

Price & Availability

The new Epson Home Cinema LS Projectors are available through Epson’s Online Store and Select Retailers. 

Related Reading:

MediaTek is taking its first proper step into desktop-style computing, with its new Dimensity CX C10 Max confirmed for an upcoming premium Googlebook.

The new chip is the first member of MediaTek’s Dimensity Compute Experiences (CX) family, and it’s built around a fairly serious set of hardware. Manufactured on TSMC’s 3nm process, the C10 Max uses an octa-core all-big-core CPU, an 11-core GPU and MediaTek’s eighth-generation NPU.

Beats’ new 360 headphones are different from any Beats pair we’ve seen before. They’re designed around an adaptable core frame with easily swappable earcups and headband cushions in multiple colors and styles, along with features like much-improved noise canceling compared with the Studio Pro and IPX4 water resistance.

Their $350 MSRP is the same as the 2023 Beats Studio Pro and considerably less than competing models from Bose, Sony, and others. But can they really be your everything headphones?

Skip the reading? Click play above or watch the Beats 360 review on YouTube.

The 360’s unique form factor takes a cue from true wireless earbuds, with a battery and circuitry in each earpiece, allowing for a thin, flexible headband. They’re still relatively heavy at nearly 300 grams, but the balanced design fits well and remained comfortable for hours of wear. The four-piece control system is simpler and more elegant than the Studio Pro’s. There are nine microphones in total; eight are used for noise canceling and transparency mode, while three contribute to excellent call quality alongside DSP and AI-powered noise suppression.

The headphones ship with diamond-knit textile earcups designed for durability, and they work reasonably well for workouts. The core frame comes in four colors: Black, Cloud, Sky, and Pink. The swappable earcups and pads add Navy and Volt Green, along with faux-leather options in white and black. Introductory pricing starts at $10, rising to as much as $50 per set afterward. Custom 40 mm drivers are powered by what Beats calls a “high dynamic range amplifier.”

Features include iOS options like Audio Sharing and hands-free Siri, while both Android and iOS users get one-touch pairing, audio switching between connected devices, and finder functionality. There are no auto-pause sensors, manual EQ, or line-in playback options of any kind. Battery life is rated at 32 hours per charge, with an hour of playback from five minutes of charging, which matched my testing. Unfortunately, the battery continued to drain after the headphones were powered down, at a rate of around 0.5 percent per hour. Beats assumed this was due to the always-on Find My feature, but the issue persisted with Find My turned off. I’m hoping Beats offers a fix soon.

Noise canceling is excellent for the money, testing just a notch or two below top models like Bose’s QC Ultra Gen 2 and Sony’s WH-1000XM6. Transparency mode is also solid, though not as clear or natural-sounding as Apple’s AirPods Pro or AirPods Max.

Sound quality is generally balanced and includes head tracking for 3D audio formats like Dolby Atmos. The bass isn’t as potent as on previous Beats headphones, including the Studio Pro, but it’s musical and well controlled. The treble is brighter than on models like Sony’s XM6, and the 360 doesn’t offer the same level of definition, but it remains relatively detailed across the frequency range, with good clarity and instrumental separation. For a closer look at performance, watch our full video review on YouTube.

The Bottom Line

The latest Beats headphones have a lot to offer, from their adaptable, splash-resistant design to powerful noise canceling and accessible sound. They’re an enticing option for folks who want a stylish, sporty pair of headphones, as well as those who don’t want to spend up for the latest flagship models from Bose, Sony, and others.

I do wish they had a few more features, including auto-pause. It would also be nice to get a hard case, though I don’t expect one from Beats at this point. The biggest issue is the odd battery drain when powered off, which makes the 360 tough to recommend based on my current testing. I’m hoping Beats can fix the problem, and it’s always possible the issue is limited to a small number of units like mine.

Otherwise, the 360’s solid performance and unique design would make them worth considering for anyone looking for versatile headphones that stand out from the crowd.

Pros:

• Distinctive customizable design
• Clear and relatively detailed sound
• Excellent noise-canceling for the money
• Fantastic call quality
• IPX4 water resistance

Cons:

• Potential battery drain issue when powered off
• Bass isn’t as big as Beats fans might require
• No manual EQ or line-in playback options
• No auto-pause

Our Ratings:

Editor’s Note: These ratings reflect the battery drain issue we experienced and will be updated if it’s fixed.

★★★★★★★★★★ Performance

★★★★★★★★★★ Build Quality

★★★★★★★★★★ Comfort

★★★★★★★★★★ Features

★★★★★★★★★★ ANC

★★★★★★★★★★ Value

Where to Buy:

Related Reading:

The most powerful AI chip in China today, delivering three times the performance of its predecessor, the Zhenwu M890.

That is how Eddie Wu described the processor Alibaba unveiled on Tuesday. The chief executive was speaking at Apsara, the company’s annual cloud conference, in Hangzhou.

The chip is the Zhenwu V900. Alibaba said in a post on X that it comes from T-Head, its chip design unit. The company says it handles both high precision model training and ultra low precision inference.

Two other announcements came with it. Alibaba plans models of up to 10 trillion parameters, and 20GW of data centre capacity by 2032.

What is inside the V900

T-Head has published a basic spec sheet for the chip. Simon Sharwood reported the numbers for The Register. The sheet lists 216GB of memory and 1,200 GB/s of inter-chip interconnect bandwidth.

The processor supports FP32 down to FP4 natively. T-Head reworked the Tensor Core arithmetic unit for better instruction precision in FP8 and FP4. The chip also carries more scaling factor formats and block size configurations under MXFP8 and MXFP4.

Wu said a single cluster can hold up to 500,000 of the chips. The resulting machine will power frontier model training and inference, he said. Alibaba launched the M890 in May.

Mass production and commercial release are set for the first quarter of 2027. Liam Mo and Eduardo Baptista reported for Reuters that Wu expects significant growth in annual AI chip shipments.

Alibaba will refresh the lineup every year, Bloomberg reported. It put deliveries of the Zhenwu family at 560,000 units to more than 400 external customers. Alibaba’s own post puts the figure higher, at more than 650 customers. They span industries from cars to language models and embodied intelligence.

The company also set out a roadmap for its next generation Yitian server processors. They are built for agentic AI work and due in 2027.

Qwen 5 is planned at 10 trillion parameters

Alibaba is already training Qwen 4. Wu named the two models after it, Qwen 4.5 and Qwen 5. Both should scale to between five and 10 trillion parameters.

Qwen 3.8 Max, the current flagship, has 2.4 trillion. Alibaba called it its most capable model in August.

Wu framed the jump as a route to artificial superintelligence. The larger models target longer and more complex tasks.

He also said the Qwen team is working on recursive self improvement. A model identifies its own weaknesses, runs experiments and generates training data with little human involvement. He described the progress as meaningful.

Twenty gigawatts by 2032

The third announcement was infrastructure. Wu set a target for Alibaba Cloud to pass 20GW of global data centre capacity by 2032.

Citigroup estimates that much capacity could drive more than $160bn of external revenue for the cloud division, according to Bloomberg.

The commercial property firm Cushman and Wakefield counts 37.7GW of data centres under construction in the United States alone. SpaceX had roughly 1.4GW of AI computing capacity in the middle of this year. It wants to pass 10GW in 2027, the Associated Press reported.

Supply is the constraint Wu named. He said global shortages across the AI data centre supply chain are limiting how fast Alibaba can scale its compute, and that mid to long term industry demand far outpaces what the company can supply.

Alibaba Cloud will start bringing its AI supernodes online at commercial scale this quarter. Wu did not say where the new data centres will be built.

What the push costs

Alibaba has committed more than $53bn over three years to its AI expansion, and raised about $10.2bn in a follow-on share sale in August.

The spending shows in the accounts. Quarterly profit fell by 75% after a spree of almost $10bn.

Executives expect annualised revenue from AI products to reach $10bn this quarter, and say the company can recoup its investment within three years. The target for cloud and AI revenue is $100bn in five years.

Investors took the announcement well. Alibaba’s Hong Kong listed shares rose 5.1% on Tuesday, to their highest in a month, and Tencent gained more than 7%.

Why Alibaba designs its own chips

United States export controls keep Nvidia’s most advanced accelerators out of China. The same controls largely block Chinese firms from using TSMC. Bloomberg reported that it is unclear whether Alibaba works with SMIC, a capacity constrained rival.

Huawei made its own move last week, unveiling new chip technology at its Connect conference.

Neil Shah, vice president at Counterpoint Research, told the AP that using extra computing power to make up for chip limits helps China stay strong locally. His colleague Parv Sharma pointed at manufacturing. How far the gap narrows also depends on China’s ability to advance its own foundries, he said, not just on chip design.

A day before the summit

Xi Jinping arrives in Washington on Wednesday for a state visit and a meeting with Donald Trump. AI, trade and tariffs should all feature. Jensen Huang and Satya Nadella are due at the state dinner.

Treasury Secretary Scott Bessent and Vice Premier He Lifeng met this week. They agreed to set up a dialogue on AI, aimed at a common understanding of goals and threats.

Washington has accused Alibaba and DeepSeek of illicitly accessing American models, a charge Beijing rejects as groundless. Anthropic accused Alibaba of distillation in June.

Dario Amodei, Anthropic’s chief executive, called this month for a slower pace of model development. Wu addressed the argument directly.

Last year, we asserted that the more capable AI becomes, the more powerful humanity will be. Today, I remain steadfast in that conviction.

He calls it machine thinking

Wu avoided the term artificial intelligence through most of his keynote. He used machine intelligence, or machine thinking, instead.

He compared the moment to the industrial revolution. Steam and combustion engines, he said, were built to do what horses and labourers already did, and machine power now drives 99.9% of the world’s physical work.

He expects the same pattern in cognition. Machines currently produce less than 3% of all thinking, by his measure. He predicted they will eventually produce more than 1,000 times the thinking of all humanity combined.

Wu also said the groundbreaking products of the era have not arrived yet. He likened AI coding to the light bulb, an early application of electricity rather than the thing that mattered most.