https://archive.ph/U7MIm

https://archive.is/0warw

cross-posted from: https://lemmygrad.ml/post/11191434

Thomson Reuters, the media company which is also a data broker, has long provided underlying personal data for Immigration and Customs Enforcement (ICE) tools, according to documents obtained by 404 Media and sources. There are also indications its data is now part of the Palantir system ICE uses to find which neighborhoods to target.

The findings draw a clearer line between Thomson Reuters’ data business—which can involve selling names, addresses, car registration information, Social Security numbers, and details on someone’s ethnicity under the brand name CLEAR—and the specific tools ICE is ingesting the data into. The news also comes after Thomson Reuters employees sent leadership a signed letter expressing their unease with the company’s ICE and Department of Homeland Security (DHS) contracts, the Minnesota Star Tribune reported last month.

“If these allegations are true, they cut directly against Thomson Reuters’ claims that its products and services are limited to fighting serious crime and are not facilitating deportations,” Emma Pullman, head of shareholder engagement and responsible investment for the B.C. General Employees’ Union (BCGEU), told 404 Media. BCGEU is a minority shareholder in Thomson Reuters and has recently engaged the company concerning its work with ICE, BCGEU said.

💡Do you work at Thomson Reuters, Palantir, or DHS? I would love to hear from you. Using a non-work device, you can message me securely on Signal at joseph.404 or send me an email at joseph@404media.co.

An internal Palantir wiki 404 Media obtained explained Homeland Security Investigations (HSI), a part of ICE that used to be focused on criminal investigations but has now shifted to immigration enforcement, used a Palantir-built system called FALCON before moving onto an HSI internal tool. A former Palantir employee has since told 404 Media Thomson Reuters’ CLEAR specifically was used in that FALCON system.

In 2025 Palantir said it became a “more mature partner to ICE” when the company started work on other systems during Trump’s mass deportation effort. That included a tool called Enhanced Leads Identification & Targeting for Enforcement, or ELITE, 404 Media revealed in January. ELITE populates a map with potential deportation targets, brings up a dossier on each person, and provides a “confidence score” on the person’s current address. An ICE official testified about using the tool before officials detained more than 30 people which lawyers have described as a “dragnet.”

Internal ICE material showed ELITE got these addresses from various sources, including government agencies such as the Department of Health and Human Services (HHS). The material also said a source of the addresses was “CLEAR.”

Two Department of Homeland Security (DHS) sources believe the material refers to Thomson Reuters’ CLEAR. “I have to think it’s the same CLEAR,” one said. 404 Media granted several sources in this story anonymity as they weren’t permitted to speak to the press about these topics.

‘ELITE’: The Palantir App ICE Uses to Find Neighborhoods to RaidInternal ICE material and testimony from an official obtained by 404 Media provides the clearest link yet between the technological infrastructure Palantir is building for ICE and the agency’s activities on the ground.404 MediaJoseph Cox

Thomson Reuters data is also mentioned in documentation about Mobile Companion, an app made by Motorola for querying license plate scans. ICE recently sent a message to all ICE Enforcement and Removal Operations (ERO) staff, who are focused on deportations specifically, about the tool, 404 Media previously reported. The material sent to ICE said users can further enhance their investigations by combining Motorola’s license plate reader network with Thomson Reuters’ data. “Thomson Reuters CLEAR combines comprehensive public and proprietary data with nationwide license plate data from Motorola Solutions’ secure shared data network to help take vehicle-involved investigations to a more precise level,” the material said.

404 Media made multiple attempts to get Thomson Reuters to comment for this story. Originally, Thomson Reuters said it would provide information “on background” over email, but then noted the background information would be material “you can use to inform your article but not attribute to Thomson Reuters.” 404 Media explained that, like many publications, “on background” to us means we could paraphrase the information and attribute it to the company. Thomson Reuters then said, “We do not agree with your definition of ‘on background’ and therefore are unable to address the misstatements we believe you may make in your story” and ultimately refused to comment.

In procurement documents available online, DHS says “CLEAR is vital to the mission-essential, time sensitive investigative work of several DHS Components as it makes it easier to locate people, assets, businesses, affiliations, and other critical facts.”

“Without this data, DHS would not be able to identify targets associated with criminal enterprise, terrorism, and immigration fraud as rapidly,” the documents add.

Those documents show CLEAR’s data can include a person’s name, address, date of birth, phone records, driver license, motor vehicle registrations, Social Security number, marital status, household information such as their household members, and details on their public social media.

In March the Minnesota Star Tribune reported it had spoken to half a dozen Thomson Reuters employees mostly based in Eagan, home to one of the company’s largest U.S. offices, and where many of the employees’ jobs relate to CLEAR. “People are worried about the role their job has played in what has happened,” one employee reportedly said, referring to Operation Metro Surge, the DHS operation focused on Minnesota in which officials killed Renée Good and Alex Pretti. The outlet reported around 180 workers sent Thomson Reuters leadership a letter expressing their unease and asking about the company’s supervision of its DHS and ICE contracts. The New York Times later reported more than 200 employees had signed the letter.

The Minnesota Star Tribune also quoted an internal Thomson Reuters message from Kevin Appold, the company’s vice president for projects and U.S. public records. “We prohibit customers from using CLEAR to identify or locate undocumented immigrants who have not committed crimes,” it said.

From 404 Media via this RSS feed

https://archive.ph/6Iiag

cross-posted from: https://lemmygrad.ml/post/11178911

Archive link: https://archive.ph/6Iiag

Chinese researchers have unveiled a new rare earth alloy so cold and efficient it could upend decades of reliance on helium-3 and send shock waves through the global race for quantum computers or ultra-sensitive detectors.

A mini-fridge built with the alloy has achieved temperatures extremely close to absolute zero using no moving parts. And it comes at a time when the US Defence Advanced Research Projects Agency (DARPA) is actively hunting for exactly such a technology.

On January 27, DARPA issued an urgent call for proposals: develop a modular, helium-3-free cooling system for next-generation quantum and defence technologies.

Less than two weeks later, the Chinese scientists answered – with a paper published in Nature.

The alloy “has the potential for mass production. The joint team has recently successfully developed a pure metal refrigeration module based on this alloy material,” the Chinese Academy of Sciences (CAS) said on its website on February 13.

“This highly efficient cooling module could offer a stable, portable cooling source for quantum chips and support major space exploration projects with a self-reliant refrigeration system,” CAS added.

“It marks a ‘China solution’ that ends dependence on helium-3.”

In physics, the lowest possible temperature is 0 Kelvin, or minus 273.15 degrees Celsius (minus 459.67 degrees Fahrenheit), a state known as “absolute zero”.

As materials approach this temperature, they exhibit radically different properties: liquid helium loses friction, mercury becomes superconductive and much cutting-edge quantum research becomes possible.

Currently, achieving such extreme low temperature primarily relies on a technique called dilution refrigeration, which requires helium-3. This stable isotope of helium is an essential resource that China largely imports. Its main sources are linked to nuclear weapons programmes in the United States and Russia, as well as civilian nuclear power plants in Canada.

---

According to a research paper published in the journal Nature on February 11, the team employed an entirely different solid-state cooling technique known as adiabatic demagnetisation refrigeration (ADR).

In simple terms, the process involves a magnetic alloy being first placed in an existing low-temperature environment. Applying a magnetic field forces the countless internal microscopic magnets to align uniformly, releasing heat that is carried away.

When the alloy is then isolated from the environment and the magnetic field is removed, the internal magnets return to a disordered state, a process that absorbs heat and further lowers the material’s own temperature.

A major hurdle in this process has been the poor thermal conductivity of traditional materials. While they could get cold themselves, they struggled to effectively cool the surrounding components.

The collaborative team from the Institute of Theoretical Physics and the Hefei Institutes of Physical Science under CAS, together with Shanghai Jiao Tong University, has discovered a new material, a rare earth compound called EuCo2Al9 (ECA). It possesses thermal conductivity similar to metal, allowing it to efficiently channel the cold outward.

“ADR using ECA has achieved a minimum temperature of 106 millikelvin, setting a new record for metallic materials. Also, at such extreme temperatures, its thermal conductivity is one to two orders of magnitude higher than traditional magnetic refrigeration materials, overcoming the key bottleneck of inefficiently extracting the cooling power,” according to the academy.

The ADR method, which eliminates the need for helium-3, is gaining traction in the academic world.

In 2024, Peking University built two “refrigerators” using this principle for quantum computing research, which have been operating stably for several months.

Lightweight portability is poised to be a key advantage of the ECA refrigeration module. This year’s Chinese government work report mentions the goal of “cultivating and developing the quantum technology industry”.

Currently, superconducting quantum computers require massive dilution refrigerators to cool their chips to sub-kelvin temperatures. In the future, a more portable refrigeration module like this could be instrumental in building smaller, more compact quantum computers.

https://archive.ph/PUJTy

cross-posted from: https://lemmygrad.ml/post/11141812

A report by official Chinese media on Wednesday unveiled two new types of anti-drone laser weapons. A developer said such laser weapon systems, supported by intelligent targeting and multi-sensor integration, demonstrate strong responsiveness and adaptability, while enabling coordinated operations within an integrated combat network.

Footage released in the CCTV News report showed how the systems countered low-altitude, slow-flying small drones simulating covert penetration on the battlefield, flying at an altitude of 50 to 80 meters, which is within the blind spot of conventional radar systems.

More advanced threats such as fiber-optic drones, capable of transmitting real-time data and video through physical links and thus resistant to electronic interference, pose additional challenges and often require physical destruction, according to the CCTV News report.

In response to such threats, multiple counter-drone systems were featured. A container-like platform identified as the “Guangjian-11E,” or Light Arrow-11E, multi-mode terminal jamming system can rapidly lock onto targets and disrupt their onboard sensing and targeting functions, the report said.

The “Guangjian-21A”, or Light Arrow-21A, system, meanwhile, operates in a “shoot-on-the-move” mode, allowing it to maintain sustained firepower while reducing exposure to counterattacks, reported CCTV News.

In terms of engagement methods, the “Guangjian-21A” specializes in “hard-kill” capabilities, generating high-density energy beams over several kilometers to penetrate drone structures and destroy internal circuits or propulsion systems within seconds. By contrast, the “Guangjian-11E” employs a soft-kill approach, using pulsed laser energy to precisely disable key components, effectively blinding drones and disrupting their reconnaissance and data transmission functions, reported CCTV News.

These complementary “soft” and “hard” kill methods address targets at different ranges and threat levels, filling gaps in each other’s operational coverage, CCTV News said.

Both systems are equipped with phased-array radar and infrared detection systems, and can interconnect via both wired and wireless links to enable real-time data sharing and precise target identification and engagement, according to the CCTV News report.

The laser systems, as next-generation equipment, feature multi-source detection and intelligent identification, enabling rapid response and adaptive targeting. Integrated into a broader combat system, the platforms can coordinate “soft” and “hard” kill methods and link with radar and electro-optical sensors to form a rapid “detect-to-destroy” chain, Zhou Shuiliang from the Aviation Industry Corporation of China was quoted by the CCTV as saying.

Previously, multiple types of anti-drone equipment were reviewed in formations at China's V-Day military parade on September 3, 2025 to mark the 80th anniversary of the victory in the Chinese People's War of Resistance Against Japanese Aggression and the World Anti-Fascist War, addressing the importance of anti-drone warfare on the modern battlefield.

Anti-drone missile and artillery systems, high-energy laser weapons, and high-power microwave weapons reviewed at the parade are a powerful "iron triangle" that can both "soft kill" and "hard destroy" unmanned aerial vehicles.

Given the widespread deployment of drones on the battlefield, countries are now prioritizing the development of efficient counter-drone systems and platforms, Wang Yunfei, a Chinese military affairs expert, told the Global Times.

The inclusion of high-energy laser and high-power microwave counter-drone systems in the parade indicates that both have demonstrated strong performance in realistic training, and suggests that China’s laser- and microwave-based counter-drone technologies are at the forefront globally.

Zhang Xuefeng, another Chinese military affairs expert, told the Global Times that high-energy laser weapons can quickly and accurately detect, aim, and track targets. Using high-energy lasers, it only takes a few seconds to shoot a drone down.

"This genre of anti-drone systems has high accuracy, minimal collateral damage, the advantage of unlimited ammunition, and very low operating costs, making it the lowest cost anti-drone equipment in a single deployment," Zhang said.

TurboQuant looks like a pretty massive deal for running local models efficiently. The core issue they are tackling is the memory bottleneck caused by the key value cache during generation. When you are doing long context inference storing all those high dimensional vectors eats up VRAM extremely fast. Traditional vector quantization helps but usually introduces memory overhead because you have to store scaling factors or constants in full precision for every small block of data. That overhead can easily add an extra bit or two per parameter which ruins the compression targets people are aiming for.

TurboQuant solves the problem by combining two clever mathematical tricks to eliminate that overhead entirely and get the cache down to 3 bits without losing accuracy. The first part is an algorithm called PolarQuant. Instead of looking at the vectors in standard cartesian coordinates it converts them into polar coordinates. This basically separates the magnitude from the direction. Because the angles map onto a fixed predictable circular grid the model no longer needs to store those dynamic bounding boxes or normalization constants that traditional methods require. That step handles the bulk of the compression to capture the main signal of the vector.

The second piece of the puzzle is where they use something called Quantized Johnson Lindenstrauss or QJL to clean up the residual error left over from the first step. QJL uses a mathematical transform to shrink that leftover error down to just a single sign bit of positive or negative one while preserving the relative distances between the data points. This acts as a mathematical error checker that fixes any bias in the attention scores. Because it only uses one bit and preserves the geometry of the space the attention mechanism can still calculate accurate logits without needing full precision data.

They tested this on open weights models like Gemma and Mistral across heavy needle in a haystack and LongBench tasks. They managed to compress the KV cache down to 3 bits with literally zero drop in accuracy and they did not even need to do any fine tuning or calibration. On top of saving a massive amount of VRAM the 4 bit version actually speeds up attention logit computation by up to 8x on H100 GPUs compared to standard 32 bit floats. This seems like a massive leap forward for anyone trying to run long context models on constrained hardware or scale up huge vector search databases.

cross-posted from: https://lemmy.ml/post/44890650

Does the old man think we can't tell this is AI-juiced? Whatever gets the Amazon book bux Powell 😭 Nonetheless, this is extremely cool. Cutting edge iridium semiconductors and other newfangled shit keeps falling off my radar because I mainly care about the energy+obligation web binding humanity together.

In the second week of February 2026, a quiet revolution landed in the pages of Science Advances. A team from Peking University’s School of Electronics, led by researcher Qiu Chenguang in collaboration with academician Peng Lianmao of the Chinese Academy of Sciences, unveiled the world’s smallest ferroelectric field-effect transistor (FeFET). Its physical gate length: just 1 nanometer. Its operating voltage: a mere 0.6 volts. Its memory performance: an on/off current ratio of up to 2 × 10⁶, programming speeds as fast as 1.6 nanoseconds, and switching energy around 0.45 fJ/μm - roughly one-tenth the best previously reported figures.

For the first time, ferroelectric memory has achieved voltage compatibility with state-of-the-art logic transistors (typically ~0.7 V). Data can now flow between memory and compute units at the same low voltage, without charge pumps, voltage converters, or the energy penalties that have plagued attempts to integrate non-volatile memory with logic at scale. The device uses metallic single-walled carbon nanotubes (m-SWCNTs) as the gate electrode, a 2D molybdenum disulphide (MoS₂) channel, a ferroelectric layer of CuInP₂S₆ (CIPS), multilayer graphene, and hexagonal boron nitride (h-BN) in a van der Waals heterostructure. The magic lies in the “nanogate” effect: the ultra-sharp 1 nm tip concentrates the electric field, creating localised intensities high enough to switch the ferroelectric polarisation well below its nominal coercive voltage, while dramatically improving capacitance coupling. Short-channel effects - the bane of conventional scaling - are rendered irrelevant.

This is not another headline-grabbing lab curiosity. It is a material foundation for the next era of computing: ultra-low-power, in-memory, edge-native intelligence that can operate at the scale of angstrom nodes without the voracious energy appetite of today’s hyperscale silicon. And when placed within the broader Chinese technology stack - 2D semiconductors, graphene-enhanced structures, triboelectric and piezoelectric nanogenerators, flexible energy storage - it becomes something far more consequential. It becomes the energetic substrate for what I have called Digital Westphalia: a global digital order grounded in nation-state sovereignty, energetic realism, and systemic abundance rather than rentier extraction and entropic financialisation. Understanding the Breakthrough in Plain Terms

Ferroelectric transistors have long promised non-volatile memory that retains data without power, switches at high speed, and consumes almost no standby energy. Their polarisation states act like tiny, permanent magnets for electrons. The problem has always been scaling and voltage. Traditional FeFETs required gate voltages above 1.5 V to flip the ferroelectric layer reliably, while logic transistors had dropped below 0.7 V. Integrating them meant wasteful voltage step-up circuits and heat. Shrinking the gate below 5 nm triggered short-channel leakage and loss of control.

The Peking team solved both with a radical redesign. By replacing a planar gate with a 1 nm-diameter carbon nanotube, they turned the gate into a nanotip lightning rod for electric fields. Simulations and measurements show field strengths inside the CIPS layer reaching 2.7 × 10⁶ V/cm at just 0.6 V applied - more than five times the material’s coercive field in a conventional geometry. Capacitance coupling efficiency soars because the tiny gate concentrates voltage drop across the ferroelectric rather than wasting it elsewhere. The result is a device that not only scales but thrives at the 1 nm limit, immune to the physics that doom conventional designs.

Reviewers noted the achievement’s elegance: voltage efficiency exceeding 125 % (operating voltage below coercive voltage), retention and endurance suitable for real-world use, and compatibility with existing CMOS processes via 3D heterogeneous integration. The implications for large-model inference, edge intelligence, wearables, and IoT are immediate. A single edge node can now perform sophisticated inference with energy budgets measured in femtojoules rather than picojoules or worse, opening the door to truly battery-free or energy-harvesting devices.

Situating the Advance in a Thermoeconomic Framework

To grasp why this matters beyond the semiconductor roadmap, we need a different lens: Systemic Exchange Value (SEV), or what I shorthand as thermoeconomics. Traditional economics treats value as subjective preference revealed in prices, with GDP as the aggregate scorecard. It abstracts away the energetic and entropic realities that actually sustain life and production. SEV starts from the opposite premise: economies are first and foremost systems of energy transformation. All use-value is ultimately embodied energy - direct and indirect - embedded in material configurations that deliver services over time. Exchange-value is the monetary claim on that embedded energy, modulated by the Energy Return on Energy Invested (EROEI) of production and use.

Three interlocking circuits define the system:

The thermodynamic circuit: real transformation of energy into ordered structures (infrastructure, devices, knowledge), inevitably producing entropy (waste heat, disorder);

The exchange-value circuit: endogenous money and financial claims that allocate claims on future available energy in potential (AEP - available energy in potential); and

The information circuit: data, algorithms, and coordination mechanisms that reduce uncertainty and thereby improve EROEI by minimising wasteful friction.

Productive systems expand systemic abundance when they increase net AEP faster than entropy degrades it. Maladaptive systems - those that channel liquidity into low-EROEI activities (speculative finance, planned obsolescence and rent extraction, for instance) - accelerate entropy, erode adaptive capacity, and eventually face collapse or forced reorganisation.

In information technologies, the dominant metric has been “performance per watt,” but SEV demands a deeper accounting: the total energetic cost of the entire lifecycle, including the embodied energy of fabrication, the operational exergy destruction (waste heat), the systemic coordination overhead, and the long-term adaptive value created. Hyperscale cloud architectures score poorly here. They concentrate enormous computational capacity in a few geographic nodes, requiring massive cooling, redundant power generation, and transcontinental data transmission. Every query to a large language model can consume energy equivalent to a household’s daily use. The EROEIu (use-phase return) looks impressive in narrow benchmarks but collapses when externalities - grid strain, water consumption, geopolitical chokepoints on undersea cables - are internalised.

The nanogate FeFET flips this script. By slashing operating voltage and enabling seamless memory-logic integration, it dramatically reduces exergy destruction at the device level. When scaled into arrays for in-memory computing, it collapses the von Neumann bottleneck, cutting data movement energy by orders of magnitude. Paired with 2D materials that can be fabricated at lower thermal budgets and with graphene or MXene-enhanced nanogenerators that harvest ambient mechanical or thermal energy, entire nodes become energetically autonomous. The information circuit now operates with far higher informational EROEI: more useful computation per joule invested, less entropy exported as heat, and greater resilience because intelligence is distributed rather than centralised.

China’s systematic investment in the full stack - from wafer-scale 2D growth to open-source AI frameworks like DeepSeek, to self-powered IoT ecosystems - is building precisely the high-EROEI infrastructure that SEV identifies as adaptive. It is creating durable use-value that compounds over decades rather than depreciating in quarters. In thermoeconomic terms, it is expanding the envelope of available energy in potential for the entire digital sphere. Digital Westphalia as the Political Expression of Thermoeconomic Realism

Digital Westphalia names the possible emerging global digital order that aligns political-information sovereignty with these energetic realities. Just as the 1648 Peace of Westphalia ended the Thirty Years’ War by enshrining territorial sovereignty and non-interference, today’s digital analogue reasserts nation-state primacy over data regimes, technical standards and infrastructure governance - while preserving interoperability through open protocols.

The old model was a de facto American imperium: hardware, software, standards and data flows routed through U.S.-controlled chokepoints (Northern Virginia data centers handling ~70 % of global traffic, SWIFT, undersea cables). This delivered rents to a handful of platforms and intelligence leverage to one state, but at the cost of universal vulnerability and entropic inefficiency. Sanctions, extraterritorial export controls, and deplatforming demonstrated the fragility. Check out Newman and Farrell’s Underground Empire for a detailed discussion of these realities.

Digital Westphalia offers an alternative: sovereign digital territories that can choose their own data localisation, governance, and ecosystem providers, yet interconnect via open-source standards (RISC-V, Linux contributions from Huawei, HarmonyOS adaptability). The nanogate breakthrough, embedded in a Chinese stack that emphasises modularity, open architectures, and energy autonomy, supplies the material base. Nations or regions can now deploy federated networks of edge intelligence without building hyperscale data centers or begging for foreign chips under export-control threat. A developing country can equip rural health posts with self-powered wearable monitors and localised diagnostic models that run inference on-device. A mid-sized power can maintain sovereign AI capabilities for agriculture, disaster response, or industrial optimisation without ceding data sovereignty or energy security.

The savings are thermoeconomic as much as fiscal. Less need for continent-spanning transmission infrastructure. Lower grid pressure. Reduced geopolitical risk premiums on energy imports for compute. Higher systemic EROEI because intelligence is co-located with the phenomena it observes and acts upon. In SEV terms, this is liquidity allocated to high-adaptive-capacity uses rather than siphoned into fictitious capital or low-EROEI consumption.

Strategic Implications in a Multipolar World

For much of the world, the choice is sharpening. The world can double down on a U.S.-centric stack that promises cutting-edge performance but delivers vendor lock-in, energy intensity, and exposure to export-control volatility. Or we can engage the emerging open, sovereign-capable ecosystem that lowers the barrier for genuine digital autonomy. The latter does not require “choosing sides”; it requires recognising that energetic and informational realism now favours distributed, interoperable sovereignty over centralised techno-feudalism.

The Peking University nanogate is one device. But it exemplifies a broader pattern: China’s willingness to invest in the thermodynamic foundations of the information age while others financialise their way toward entropy. The full stack - 2D materials scaling, nanogenerators, open-source models, distributed ledger coordination - is creating the possibility of Digital Westphalia at planetary scale. Nations that seize it will expand their adaptive capacity; those that cling to the old imperium risk locking themselves into maladaptive rigidity.

This is the real significance of the 1 nm breakthrough. It is not merely smaller and lower-power. It is a material refutation of the assumption that computational abundance must come at the price of energetic profligacy and political subordination. In thermoeconomic terms, it augments available energy in potential. In geopolitical terms, it makes Digital Westphalia not utopian but all but inevitable.

The question for policymakers, strategists and citizens is no longer whether a new digital order is coming. It is whether we will shape its emergence in alignment with energetic realism and sovereign dignity or allow entropic forces to dictate the terms.

Hello Lemmians,

I've recently moved over to Linux Mint for Windows and I'm very happy with the decision. There are only like 2 things I can't do on Mint but I'm confident I could do them if I invested some more time. Anyway, my system is on dual boot so I could start Windows if I wanted to / needed it.

Now, I'd love to do the same on my phone, basically eliminating techbrofash stuff and replacing it with something FOSSy. It's a Samsung from 2018, do I have any chances? Specifically, having to use a Google account all the time bugs me.

I'm not a heavy phone user since I prefer computers, but as you all understand I'm not keen on having Netanyahu & friends know at what time of night I pee.

Any suggestions?

Thanks!