News
- Chris Malone, OpenAI's head of data centers, has left after seventeen months, with the company confirming the exit but giving no reason
- His responsibilities had already been split before he went, with his reporting line moved to VP Sachin Katti and the job now spread across three executives plus an unnamed CTO for computing capacity
- The move comes as OpenAI is reportedly reviving plans to lease entire facilities rather than build them, a shift from construction to procurement
Chris Malone, the executive who ran OpenAI's data center buildout, has left the company. The Wall Street Journal broke the story last week, and an OpenAI spokesperson confirmed the departure to Bloomberg the same day.
Neither OpenAI nor Malone has given a reason for his departure or his next destination, and a message from Bloomberg to his LinkedIn account received no reply.
The top-level executive joined the company in March 2025, against the backdrop of OpenAI's Stargate announcement in January of the same year, a commitment to invest $500 billion in 10 gigawatts of US AI infrastructure over four years alongside SoftBank, Oracle and MGX.
A job cannibalized by changing priorities at the helm?With a total tenure of 17 months, a relatively short duration for the job, Malone arrived at a time when the job was incredibly demanding, as OpenAI fended off competition from rival Anthropic, a resurgent Google with its Gemini offerings, and what can perhaps be described as a love-hate relationship with Microsoft.
His exit may have been written on the wall weeks ago, according to the Wall Street Journal, which noted that Malone stopped reporting to OpenAI president Greg Brockman and began reporting to vice president Sachin Katti, who took over leadership of the group.
OpenAI has since appointed a chief technology officer to lead computing capacity, though it has not publicly named that person. Responsibility for what Malone did is now spread across at least three people: Uday Ruddarraju leads the data center team, Brent Mayo runs the build-and-delivery program, and Spas Lazarov, who came from the data center and energy industries, leads data center engineering.
All of this indicates it wasn't a sudden vacancy to fill, but one that had been on the cards for some time. This comes as reports indicate OpenAI is reviving data center initiatives involving leasing entire facilities, with others leading that effort. This would mark a significant shift from its initial plans to get involved in the building side of the business, which it seems to have focused less on lately, and a major strategic shift in how the organization plans to operate in the future.
OpenAI's own narrative points in the same direction, without conceding much, when it told TechCrunch it had "recently reorganized" its "infrastructure organization to support the scale and pace of our work," adding that it retains a strong and deeply experienced data center team with clear leadership. It, however, has not addressed Malone's exit directly.
The move is not isolated, as OpenAI continues to lose high-level executives; Business Insider tallied thirteen senior departures at OpenAI in 2026 as of August 21, and Malone is the fourteenth to leave the company to date. The Next Web counts seven senior departures or replacements since April alone.
Malone is also among the least senior names on those lists by title. The Head of data centers is not a named C-suite role, and after the reorganization, he reported to a vice president rather than sitting alongside a chief operating officer or a chief revenue officer for OpenAI.
Axios reported earlier this month that OpenAI's recent changes are "driven by strategy more than the revolving door" that characterizes the wider AI talent market, with OpenAI president Brockman deliberately building a leadership team to compete in the enterprise space. Brockman has denied this, telling the Journal that the spotlight on his company means "every departure gets scrutinized in a way that it doesn't otherwise."
The scrutiny comes with the ambitions OpenAI currently harbors, many of which seriously affect its direct partners, including Nvidia, AMD, and Microsoft. CNBC reports OpenAI is targeting roughly $600 billion in compute spending by 2030. It is planning a $30 billion data center in Georgia. Nvidia agreed on August 17 to provide up to $105 billion in financing for an OpenAI-leased campus in Ohio, after earlier reports said it was weighing a considerably larger guarantee.
The move comes at a time when the company is getting ready for an IPO that would inject fresh cash while also allowing employees to essentially cash in on their equity of what is one of the world's most valuable startups: OpenAI filed confidentially with the SEC in June, and chief financial officer Sarah Friar told staff last week the company will go public in 2027 or sooner, having originally targeted this year.
An exodus of key employees, including Malone, might just be a strategic change in how the company operates or otherwise, but given the sheer scale of the company involved and the potential consequences of such moves, the scrutiny is well-warranted in an industry that has otherwise progressed relatively unchecked in the name of progress over the past few years.
- McKesson confirmed ShinyHunters breached its Snowflake and Salesforce, stealing 284M patient records
- Data includes names, contact info, SSNs, and health details; ransom demand was $55.2M
- Boston Scientific removed attackers but faces CRM device activation issues; attribution not confirmed
Last week, two major healthcare organizations suffered highly disruptive cyberattacks: Boston Scientific, and McKesson. We now have more details about both those attacks, and it seems at least one is the work of the infamous ShinyHunters extortion group.
McKesson confirmed having been struck by ShinyHunters, just a few days after the threat actor claimed responsibility. The group told The Register they broke into the company’s Snowflake and Salesforce instances and stole “millions of patients’ data”.
The company later issued a statement, saying the stolen data belonged to its Oncology & Multispecialty and Medical-Surgical business units. A spokesperson told The Register multiple employees were targeted with a vishing attack.
Boston Scientific works on restoring systemsThe group told the publication it stole more than 284 million records of patient data and demanded $55.2 million from the victims. They are saying the stolen batch includes patient tames, postal and email addresses, phone numbers, Social Security numbers (SSN), and details regarding their health condition. Whether the claims are true, and to what extent, remains to be seen after the investigation.
Boston Scientific, on the other hand, said it successfully removed the attackers from its infrastructure, but added that the investigation into the attack remains ongoing. It also said that new Cardiac Rhythm Management (CRM) devices, implanted after August 25, cannot be activated, and the data they generate will not automatically be transmitted to remote patient management systems.
“Newly implanted ICMs (insertable cardiac monitors) must be activated using the Boston Scientific Clinic Assistant app to enable the ICM to properly record episodes,” it explained. “New ICMs are unable to pair to the patient remote monitoring mobile phone, therefore available episode data recorded by the ICM will NOT be transmitted to the remote monitoring system until the ICM can be paired to the patient mobile app. Episodes will continue to be recorded by the ICM and can be transmitted to the remote monitoring system via an in-person interrogation with the Clinic Assistant app by selecting the “Interrogate” button.”
Boston Scientific is yet to name ShinyHunters as the perpetrators, and the group has not yet publicly claimed responsibility for the attack.
Via The Register
- Supermicro's new workstation costs nearly $93,000 before any tax or shipping
- It carries 748GB of memory split across CPU and GPU pools
- The system delivers up to 20 PFLOPS of FP4 performance overall
Supermicro has begun selling a workstation built around Nvidia's GB300 Grace Blackwell Ultra superchip, priced at $92,887.47 before tax and shipping costs.
That figure makes the machine roughly seven times costlier than a comparably specced Mac Studio running Apple's M5 Ultra chip, which retails at $12,299.00.
The total memory on the Supermicro workstation reaches 748GB, compared with just 256GB on Apple's Mac Studio with an M5 Ultra chip.
Built for heavy AI workloadsNvidia's platform instead splits its 748GB total between 496GB of system memory attached to the Grace CPU and 252GB of HBM3e memory sitting on the Blackwell Ultra GPU.
Despite that structural difference, the combined pool still gives buyers nearly three times the addressable memory found in the Apple alternative.
The workstation pairs a 72-core Grace processor with a Blackwell Ultra GPU, forming a coherent memory architecture Nvidia says supports trillion-parameter models locally.
Storage includes two 1.92TB and two 960GB M.2 NVMe drives, giving buyers a flexible mix of fast local capacity.
Additional M.2 slots support PCIe 5.0 and PCIe 6.0 standards, allowing further expansion for datasets used in extended research work.
Networking runs through dual 400GbE ports paired with Nvidia's ConnectX-8 SuperNIC, alongside a dedicated management LAN port for remote administration.
For connectivity, the system includes several USB ports, a mini display output, and IPMI support for remote diagnostics and monitoring.
Nvidia claims the platform delivers up to 20 PFLOPS of FP4 performance, tripling inference speeds compared with its H200 chip.
The system also supports continuous agent workloads, generating as much as 1200 tokens per second at no additional token cost.
Cooling relies on a direct-to-chip liquid system, paired with several fans mounted inside a tower or rack-mount chassis.
Its power draw comes from a single 1600W supply rated at 94% Titanium-level efficiency for sustained heavy operation.
The chassis weighs about 88 pounds unloaded, reflecting how densely its components are packed into a compact frame.
Pricing puts it far beyond consumer reachSupermicro built the workstation for research labs and enterprises running large models locally, away from typical home or office computing needs.
Apple's Mac Studio, meanwhile, serves creative professionals and general users who need strong performance without specialized AI infrastructure at home.
Buyers seeking lower entry costs may find a 3% discount available for new customers on the Gold Series model.
That discount brings the price down to roughly $90,101.05, still far above what most buyers would consider for a desktop workstation.
Volume buyers can also request quotes for larger discounts, a common practice among vendors selling into research and laboratory settings.
Supermicro lists the workstation under model number ARS-511GD-NB-LCC-01-G2, part of its Gold Series lineup aimed at research labs.
For most buyers, a device like this only makes sense if trillion-parameter models must run entirely on-site, without cloud reliance.
Otherwise, the price gap compared with prosumer alternatives seems difficult to justify for typical research budgets outside well-funded organizations.
- X-Bow received nearly $11 million under a new interceptor development contract
- The new interceptor system targets a cost below $750,000 per launch
- Patriot missiles currently cost roughly one-fifth more than this new target
The Pentagon has awarded X-Bow Systems almost $11 million to develop a lower-cost interceptor system under a missile defense program.
The project falls under the Missile Defense Agency's Rapid Response Small Launcher Technology program and seeks to create a weapon that could cost below $750,000 per launch.
That figure amounts to roughly one-fifth the cost of a Patriot missile, the current top-tier standard for air defense systems.
X-Bow develops a new interceptor approachThe Missile Defense Agency selected X-Bow through its Low-Cost Interceptor program, which focuses on developing affordable systems against advanced aerial threats.
The company will use the funding to continue propulsion work and prepare a prototype for a planned flight demonstration.
X-Bow produces its solid rocket motors and propellant internally, allowing the company to control more parts of its manufacturing process.
The agreement requires the firm to deliver a prototype system by late 2027, although that stage will not represent a fully operational weapon.
The Pentagon has faced pressure because existing interceptor systems remain expensive and difficult to produce quickly.
The war in Iran revealed a major setback as the U.S. is now left with about 250 THAAD rounds and less than 1,000 Patriot interceptors.
Patriot replacement rounds cost around $3.9 million each, while manufacturing takes years and depends on a small number of established suppliers.
“The contract will aim to address the critical need to expand the SRM industrial base, which is currently reliant on a duopoly of two primary manufacturers,” said the US Department of Defense in its contract announcement.
X-Bow said its production methods could help create affordable weapons while increasing manufacturing speed for military requirements.
“America can’t out-produce today’s threats with yesterday’s supplier base,” said Jason Hundley, founder and CEO of X-Bow Systems.
The company is examining ways to reduce costs across interceptor components, including propulsion systems and other expensive sections.
Other nations explore cheaper missile defense systemsThe search for lower-cost interceptors has expanded beyond the United States as allies face similar concerns.
Ukraine and European partners have started developing alternative approaches that could reduce reliance on existing expensive missile defense weapons.
One European effort called Freyja has been linked with an interceptor costing around $700,000 per round.
Lockheed Martin has also introduced the PAC-3 ACE missile concept, which the company says could lower costs compared with current Patriot versions.
The US Air Force previously requested proposals for a counter-air missile that could cost below $500,000.
That effort reflected wider interest in producing larger numbers of affordable weapons for future defense requirements.
The X-Bow project remains in an early development stage because testing and evaluation are still required before deployment.
Military officials will need to confirm whether the system can achieve lower costs while meeting operational standards.
“We’re taking a look at propulsion, we’re taking a look at all the expensive parts of an interceptor to look for opportunities to drive down every bit of cost,” said Lt. Gen. Heath Collins, director of the Missile Defense Agency.
The program shows the Pentagon is exploring alternatives as demand for missile defense grows worldwide.
However, the outcome will depend on whether new production methods can deliver affordable systems without reducing expected performance levels.
Via Defense News
- Anthropic wants AI agents to control programmable machines through common software rules
- Model Hardware Standard (MHS) can connect microscopes, robotic arms, and liquid handling equipment
- MHS allows different laboratory devices to communicate through standardized software drivers
Anthropic opened a preview of its Model Hardware Standard (MHS), designed to let AI agents control programmable machines through shared rules.
The proposed system is intended for laboratories, factories, and workplaces where different instruments must work together without custom software links.
MHS uses software drivers that translate between computer systems and individual machines, allowing equipment with different interfaces to communicate consistently.
A common system for physical machinesIts drivers rely on basic commands for reading measurements or changing settings, giving agents a method for controlling connected equipment.
The system records machine details that might otherwise remain inside manuals or depend on knowledge held by experienced laboratory staff.
Users can enter such information through natural language, after which MHS creates reference material describing capabilities, adjustments and safety restrictions.
Agents can then discover compatible equipment across networks without requiring separate software bridges for every machine they need to operate.
The standard supports microscopes, liquid handlers, robotic arms and other equipment that offers some form of programmable interface for agents.
The company claims integration that once required weeks or months could instead take hours or minutes when equipment supports MHS.
According to Anthropic, Claude was tested with physical experiments, including laser alignment supported by camera-based observation during early physical trials.
The model successfully adjusted the laser, checked the image, and repeated the process while assessing each change carefully.
The design also lets agents combine commands from several devices, creating automated sequences for work that might otherwise require coordination.
Physical limits remain a concernAnthropic said Claude interacts with experiments and hardware in an “exploratory manner, much as a scientist would,” during physical testing.
That model can then coordinate several instruments through one interface, rather than requiring separate control software for every connected device.
The company has shared MHS with manufacturers and research groups working across biotechnology, robotics and quantum computing fields.
Anthropic intends the preview to help establish safety checks before the standard becomes available widely to developers and equipment makers.
Amazon Web Services, Automata, Danaher, Doosan Robotics, MBF Bioscience, QIAGEN, Tecan and Universal Robots are among the early participants now.
Anthropic acknowledges that AI agents still require expert supervision because language models have clear limits when reasoning about physical equipment.
MHS cannot currently support equipment lacking a programmable interface, leaving some laboratory and industrial machines outside its current technical scope.
Anthropic says it plans to work with manufacturers to add drivers while expanding compatibility with more devices and robotics platforms.
The preview will also support new safety tests and practical rules for deploying AI agents around physical equipment.
The company says its eventual open-source release will include findings from these tests alongside detailed guidance for safer implementation.
The project therefore remains experimental, with practical limits still depending heavily on hardware design, software access and human oversight.
A group of developers reportedly lost their jobs during what their employer called an “AI Transformation.” Their response was unusually literal. They built OpenExecutive, an open-source system designed to perform work associated with a CEO and the rest of the executive suite.
The employer and CEO have not been named in the online discussion, and OpenExecutive’s own materials don't present it as revenge for a particular firing. That said, it's hard not to see some specific grievances in both the project and the response online, with lots of comments gleefully welcoming senior management to the automation discussion.
The appeal is easy to understand after several years of executives explaining that AI will make organizations leaner, usually while standing a comfortable distance from the part being trimmed.
OpenExecutive presents one AI executive to the user, but that voice is backed by eight specialist agents. The virtual management team includes equivalents of everything from a chief strategy officer and chief operating officer to general counsel and a board communications director. It can draw on company documents, remember previous decisions, and surface follow-up actions later.
Org chart forkedOpenExecutive uses Anthropic’s Claude models by default, with an executive agent routing requests to whichever specialists seem relevant. The developers pitch it as a C-suite group chat in which everyone responds promptly. The project is distributed under the permissive Apache 2.0 license. Anyone can inspect it, modify it, run it internally or build a commercial product around it.
That open-source element gives the project more weight than a chatbot sold as a digital chief executive might usually represent. It also allows workers to experiment with the same logic their employers have been applying to them. If a company can divide software development into tasks, feed those tasks to models, and keep a few people around for supervision, it can do so.
The comparison is not entirely frivolous. Executive work involves processing information, preparing scenarios, comparing trade-offs, and producing documents for other executives. Those are all activities developers think AI models can handle. It makes OpenExecutive both a product and a satire of AI products.
AI replacement has often been discussed as something that moves upward from the bottom of an organization, consuming routine work before eventually reaching management. The developers have turned the telescope around and discovered that a surprising amount of executive labor also looks like information moving between boxes.
Human accountabilityOpenExecutive is still an executive adviser, not a legal chief executive. It cannot accept responsibility or sign documents on behalf of a company. A human must decide whether to act on its recommendations and remain accountable if those recommendations fail.
That's where human leadership can't really be replaced. Analysis is part of the job, but it also involves making choices under uncertainty and accepting responsibility for the consequences. A model can recommend closing an office or cutting a team without feeling the weight of the people affected, while the humans deploying it may find the machine’s apparent neutrality rather convenient.
The project therefore reveals one of the stranger assumptions behind AI-driven layoffs. Companies often judge workers by how much of their output software can reproduce. Executives are more often judged on relationships and judgment.
AI does not erase that distinction, but it makes the special exemption harder to defend. If a developer must prove that their work contains something beyond the code an AI can generate, senior leaders should be expected to prove that their value exceeds AI recommendations. A large salary should buy more than the ability to schedule another meeting about efficiency.
Recent corporate experiments also suggest that replacing people can be easier to announce than to sustain. For instance, Duolingo’s CEO softened an “AI-first” memo after backlash over plans to phase out contractors where automation could handle their work.
Automation keeps happening, but there should be more questioning of its benefits. AI can help a small group start a company without immediately assembling a miniature court of consultants and executives. It can also give existing leaders a new justification for concentrating control while reducing the workforce beneath them.
OpenExecutive will not make CEOs obsolete, at least today. Its recommendations require scrutiny and auditing. But it shows the rhetoric of AI efficiency goes both ways in corporations. Once leaders decide that every job should be broken into tasks and measured against a machine, their own work becomes fair material for the same experiment. The resulting AI CEO may never fire its human counterpart, but it has already made the performance review considerably more awkward.
- TANK 5 Pro covers a 95-degree field with its thermal sensor
- The battery holds 17,600mAh and recharges fully within 90 minutes
- Dimensity 9400e reportedly outscores Snapdragon 8 Gen 3 on AnTuTu benchmarks
8849, a Chinese company known for producing rugged smartphones, has introduced the TANK 5 Pro, a device engineered for outdoor and industrial use cases.
This smartphone measures 177.1 x 87 x 33.8mm and weighs 715g, a build that supports its shockproof, reinforced construction.
A built-in toolbox app adds a compass, bubble level, protractor, barometer, and speedometer, tools aimed at fieldwork rather than everyday convenience.
Rugged design, flagship chip, and a 2k projectorThe 8849 TANK 5 Pro includes a 1200-lumen camping light with SOS and strobe modes, alongside red and blue warning lights meant for roadside or rescue situations.
Inside sits a MediaTek Dimensity 9400e processor on a 4nm process, which the seller claims scores above 2.3 million on AnTuTu, edging past the Snapdragon 8 Gen 3, which scores about 2.2 million.
It uses a 6.73-inch AMOLED panel supporting 1440×3200 resolution with a 120Hz refresh rate and 3000-nit peak brightness.
Memory options include 18GB of physical RAM plus an additional 18GB of virtual allocation, alongside 512GB of storage expandable via an microSD card.
A 17,600mAh dual-cell battery anchors this rugged phone, supporting 120W wired charging that the listing says fills the cell in roughly 90 minutes.
This mobile phone also comes with a 2K 220-lumen projector with laser autofocus and ±45° keystone correction for distortion-free setup in varied lighting conditions.
The same laser module doubles as a four-meter rangefinder, while the projector lens can also serve as a high-intensity flashlight.
It supports IP68 and IP69K certification under the IEC 60529 standard, tested in static fresh water up to 1.8 meters deep for as long as 30 minutes.
8849 cautioned that seawater, pool water, soapy water, and hot water can all reduce that protective performance over time.
An ultra-wide 'predator' thermal camera behind the lensThe infrared module, nicknamed for its wide, predator-style field of view, covers 95° horizontally, nearly double the 50° span found on most rival thermal phones.
It uses a FLIR Lepton 3.1R sensor with 160×120 resolution, translating to 19,200 individually calibrated temperature-reading pixels rather than a simple heat-blob overlay.
Sensitivity is rated below 50 millikelvin, meaning it can register shifts as small as 0.05°C, a spec aimed at building inspectors, electricians, and search-and-rescue teams.
Two gain modes stretch the measurable range from −10°C up to 400°C, covering tasks from cold-storage checks to industrial machinery diagnostics.
FLIR's own MSX and Vivid-IR processing sit on top of the sensor, sharpening edges and reducing blur in the thermal feed.
A 50MP main sensor, 50MP telephoto lens, and separate 50MP night vision camera round out the rear array, supported by four infrared LEDs for low-light capture.
A 32MP front-facing camera completes the setup, handling calls and selfies with fixed-focus clarity.
None of these individual specs is unheard of on its own, but bundling a wide-view thermal sensor, projection, and flagship-tier processing into one rugged shell is unusual.
In terms of its durability, the company listed some caveats, including the variability of drop resistance depending on impact angle and surface hardness. It therefore recommends the original case for added durability.
Water, dust, and shock resistance are described as non-permanent, gradually declining through daily wear rather than remaining fixed for the device's lifespan.
Recommended operating temperature sits between 0°C and 35°C, with prolonged ultraviolet exposure noted as a factor that can affect the housing over time.
The phone is listed at £903.01 (about $1,224), with the seller citing rising memory, chip, and tariff costs as the reason for no launch discount.
Can you believe that the Dyson Supersonic turned 10 in April this year? When it arrived back in 2016 we were all amazed at its feat of engineering, with Dyson using its expertise in airflow to create a hair dryer that styles and dries hair quickly, efficiently and with minimal heat damage.
Since then, though, the original Supersonic has spawned new models, such as the Supersonic Nural and Supersonic r, but for simple, everyday drying, the OG is tough to beat. And while it still commands Dyson money, the Supersonic Origin is at the receiving end of a 45% price cut at David Jones and can be yours for AU$299.
The Supersonic Origin might only be the bare bones model of Dyson’s famed hair dryer, coming supplied with a single attachment in the styling concentrator, but it’s still more than capable of quickly drying and styling hair from its ergonomically designed body. View Deal
In our original Dyson Supersonic review, we were thoroughly impressed with its ability to try long, relatively thick hair — it managed to fully remove moisture from towel-dried hair in just under four minutes on the lowest heat setting and highest fan speed. Increasing the temperature and lowering the fan speed resulted in a quicker two and a half minutes.
On that note, the Supersonic offers three fan speeds and three temperature settings, giving you a good amount of control and the ability to prevent heat damage. We also loved how it leaves hair silky smooth.
Of course, the ‘standard’ Dyson Supersonic comes with extra attachments compared to the Origin model on sale, which are designed to be used with specific hair types or use cases — a diffuser for curly hair or a styling nozzle for straightening, for example. If you know you’d benefit from extra attachments, you can still buy them from Dyson directly with the cash you’ll save from this deal.
I’ve seen the Dyson Supersonic Origin drop to AU$299 before, but that was last year during the Black Friday sales and only as an exclusive deal on the company’s eBay store. That means this monster saving at David Jones is one to consider if you’ve always wanted the capable hair dryer, but haven’t wanted to splurge at full price.
Please note that while David Jones lists the RRP as AU$499, Dyson lists it at AU$549, making this a truly stunning discount.
More Dyson dealsIf it's a Dyson hair straightener you seek, the Airstrait works wonders. It can be used on wet hair without causing damage to not only dry and straighten hair, but style it too.
Also available in Bright nickel/Rich copper and Amber Silk for the same price.View Deal
The V8 Cyclone is a fantastic stick vacuum for everyday cleaning. It delivers more powerful suction than the regular V8, a longer runtime of up to 60 minutes and 3 suction modes. View Deal
V16 Piston Animal: was $1399 now $999
Dyson’s most powerful vacuum to date not only benefits from exceptional suction power but also incorporates a few useful design additions, such as a lever to squish dust in the canister to help extend cleaning sessions (and to easily push it out when you’re done), and a cuff at the top of the wand that releases nozzle attachments without you having to bend down. It also gets the company’s All Floors Cones Sense floorhead to help prevent tangled hair.View Deal