CONNECT WITH US
Friday 4 July 2025
Yo-Kai reinvents late-night Asian food with smart vending machines
Numerous restaurants in American cities have already closed their doors as twilight descends. Typically, the only options available to you if you are currently famished are microwave meals or fast food. This life pain point led Andy Lin, founder of Yo-Kai Express, to start a business and design a smart vending machine capable of delivering high-quality hot cuisine at any time and from any location.Lin, who worked for a leading technology company during his early career and lived in Silicon Valley for an extended duration, frequently conducted meetings with Asia clients late at night in the US. He was so hungry that he searched the entire Internet one early morning, but he was unable to find a venue that served a steaming bowl of Asian cuisine. Yo-Kai was established as a result of this experience.Lin initially drew inspiration from Japan's pioneering ramen dispensing machines; however, he ultimately determined that the user experience was inadequate. He utilized his technological proficiency to reengineer the mechanism and process, resulting in the creation of a smart heated food vending machine that is "delectable, sanitary, and prepared for consumption."Expand product lines and explore instant food blue oceanYo-Kai's first product is a ramen machine that uses high-pressure heating and quick cooking technologies to deliver restaurant-quality dishes. All components, including the broth, noodles, and seasonings, are crafted from premium ingredients that are rapidly frozen and pre-prepared in a central kitchen to guarantee uniform quality.Lin recalled that in his pursuit of the ideal noodle flavor, he once produced 3,000kg of noodles in a single day, only to meticulously control the variance of a few seconds. Yo-Kai's product selection now includes Korean, Thai, xiaolongbao, chicken noodle soup, and several other items, and has collaborated with companies such as Ippudo, MoonMoonFood, Chimei, and Wowprime to deliver joint products.Yo-Kai's smart vending machines are capable of both cognitive processes and food preparation. Sophisticated tasks, such as image recognition, virtual assistance, and sensor monitoring, are facilitated by the industrial computer module, which was developed in partnership with Qualcomm and Pegatron. The device is equipped with more than 100 integrated sensors that accurately measure parameters such as pressure, temperature, and humidity, thereby guaranteeing consistent food preparation.What is even more remarkable is that Yo-Kai utilizes edge AI processors to virtualize customer service and operation tutorials, which are then managed in real time on the spot. Users can obtain assistance without the need to make a phone call or connect to the cloud, which will further enhance the security and convenience of their information.Yo-Kai was the first-place winner of the Qualcomm QITC AI competition in 2024 and was chosen by Nvidia to participate in the Inception Program in 2025. It intends to become the Amazon of the catering industry by integrating upstream and downstream using AI systems.Alongside hot meals, Yo-Kai has lately diversified into cold beverages with better gross profit margins and has adopted an innovative approach by introducing a smart pearl milk tea machine. This is the sole vending machine globally capable of autonomously boiling tapioca pearls. It addresses the challenge of cold chain logistics for conventional tea beverages. The items can be maintained at ambient temperature, facilitating the improvement of delivery networks.Lin forecast that the tea beverage market in Europe and the US will reach US$40 billion by 2030, with an annual compound growth rate of 13%. Although vending machines have the potential for significant future expansion, their current prevalence is negligible. He intends to employ this as a secondary growth generator for Yo-Kai revenue.Yo-Kai has acquired operational experience and data over the past eight years by employing a direct sales model. The global franchise plan will be officially promoted beginning in 2025. For instance, each machine in the US is priced at a mere US$2,000-2,500. In order to operate a smart restaurant, the franchisee is only required to be responsible for replenishing stocks and cleaning, rather than preparing and opening the store. Lin disclosed that the global franchise is making progress ahead of schedule. Nearly 500 smart pearl milk tea machines have been signed in the first four months, and the company has expanded to the US, Japan, the UK, Singapore, Thailand, the Netherlands, and France, among other countries.TTA facilitates international connections, enabling Yo-Kai to expand globallyThe first smart food vending machine is being deployed worldwide. During the business process, Lin regularly returned to Taiwan to interact with various Taiwanese companies. He noted that in recent years, Taiwan's start-ups have increasingly ventured beyond the local domain, significantly enhancing their international presence. Multiple teams have actively established facilities in Singapore, Japan, and other regions to expand their global outlook. Lin contends that TTA (Taiwan Tech Arena) has played a crucial part in this transformation. Yo-Kai demonstrates exceptional industry engagement by utilizing global resources through TTA, connecting Taiwan with foreign assets, and participating in CES exhibitions on many occasions.Beginning with the resolution of the minor issue of "lack of hot food during late hours," Yo-Kai has gradually developed an advanced vending machine system that combines food, technology, and supply chain logistics. Lin revolutionizes the catering sector with technology innovation and reimagines ready-to-eat meals with uniqueness and execution. In the future, we may be able to have a cup of hot ramen soup from Yo-Kai in a variety of locations throughout the world, including airports, hospitals, and campuses, while feeling the comforting flavor offered by technology.
Thursday 3 July 2025
A-TOP Health Biotech unveils JOUVENCE thread lift, expands footprint in global medical aesthetics market
The self-funded medical beauty market has gradually integrated into consumers' daily life, combining the pursuit of outer aesthetics with inner self-confidence, resulting in significant growth in the worldwide medical beauty business.According to a June 2024 survey conducted by the International Society of Aesthetic Plastic Surgery (ISAPS), plastic surgeons worldwide conducted over 15.8 million surgical procedures in 2023, in addition to the 19.1 million non-surgical procedures. Among them, non-surgical skin tightening treatments were ranked fourth in market demand.To capitalize on the substantial business potential within this sector, A-TOP Health Biotech, possessing robust R&D capabilities, introduced its proprietary product "JOUVENCE" to address the shortcomings of current thread materials. The product will soon be submitted for review by the U.S. FDA, with certification expected in the fourth quarter of 2025.As emphasized by Olivia Tseng, founder and chairperson of A-TOP Health Biotech, ongoing advancements in medical aesthetics technology have made major surgical procedures unnecessary for achieving skin firmness. Instead, lifting threads can now be inserted into targeted areas to achieve the desired results. Conventional lifting threads feature a "sharp barb" or "hooked" design that pierces the skin tissue, granting anchorage to facilitate skin lifting; however, this design often causes discomfort, and in some cases, patients may experience excessive inflammation requiring additional surgery.JOUVENCE lifting threads adopt a unique bionic cross-sectional design to avoid the discomfort caused by conventional barbed threads, thereby offering consumers a more comfortable and effective treatment experience.Tseng stated, "JOUVENCE will challenge the traditional paradigm of round thread materials, redefine the technical concept of thread lifting, seamlessly integrate technology and aesthetics to usher in a new era of innovation, and establish itself as a pivotal and trusted brand in the global medical beauty sector."3D Beauty-AI software enables innovation in lifting threadsA-TOP Health Biotech, founded in 2021, is driven by the mission of "Technology and Beauty, Touching the World". The company focuses on developing medical aesthetic and reconstructive surgical materials that are both clinically proven and highly safe, aiming to solve clinical challenges and enhance the quality and efficiency of aesthetic medicine.Its flagship product, JOUVENCE, is a world-first innovation developed using patented non-cutting molding and twisting technology, overcoming the limitations of traditional PDO lifting threads and offering four major advantages. First, its continuous structure provides stable support, reducing the risk of thread slippage and displacement. Second, it eliminates the breakage issues caused by stress concentration in cut threads, resulting in smoother surgical procedures. Third, the absence of sharp structures minimizes the sensation of foreign objects and reduces tissue irritation, significantly lowering inflammation and fibrosis. Fourth, it complies with standardized certification protocols, demonstrating rigorous biocompatibility and excellent product performance.In addition to product innovation, A-TOP Health Biotech addresses clinical pain points by introducing its "3D Beauty-AI Software" to assist physicians. By processing just 15 facial images, the software completes a 3D visual reconstruction, helping non-specialist doctors quickly grasp the key techniques for thread lifting and shortening the learning curve. The AI-generated 3D images also facilitate doctor-patient communication, enhancing transparency and trust in treatment, and further improving the overall medical aesthetic experience.Enhancing worldwide patent deployment: Focus on Taiwan, Thailand, and the USDespite Tseng's extensive practical experience in the medical devices and biotechnology sectors, A-TOP Health Biotech may continue to face multifaceted challenges in areas such as R&D, marketing, and fundraising as the company expands.By joining the Taiwan Tech Arena (TTA), the company gained valuable resources and received comprehensive support from the FlyingVest Accelerator, gradually overcoming key early-stage startup obstacles. The company also successfully secured a spot in the MEDICA Taiwan team organized by Taiwan's National Science and Technology Council (NSTC) in 2024, demonstrating both its technological strength and market potential.In the context of global strategy, Tseng stated that A-TOP Health Biotech will prioritize Thailand, Taiwan, and the US as target markets, combining its JOUVENCE lifting threat technology with 3D Beauty-AI software to provide medical beauty clinics with comprehensive non-surgical skin tightening solutions.A-TOP Health Biotech aspires to establish itself as a globally influential medical aesthetics brand through innovative R&D and international patent strategy, realizing the brand vision of "Technology and Beauty, Touching the World."
Thursday 3 July 2025
Blockchain infrastructure driving semiconductor demand
TSMC's fourth-quarter profits jumped 57% to $11.4 billion, and while most analysts credit AI chip demand, there's another force quietly reshaping semiconductor markets. Blockchain infrastructure is creating demand patterns that extend far beyond cryptocurrency mining - we're seeing measurable pressure across multiple chip categories that most observers haven't connected yet.The applications driving this demand aren't limited to financial networks. Everything from supply chain verification systems to new bitcoin casinos requires specialized hardware that traditional computing applications never needed. Gartner projects this intersection will generate a $1.5 billion market opportunity by 2028, while Deloitte's research shows 70% of semiconductor executives expect blockchain to significantly impact their operations within five years. That's not speculation - it's capital allocation planning based on technical requirements they're already seeing.When Cryptography Meets SiliconHere's what makes blockchain different from typical computing workloads: the cryptographic operations require hardware that can handle intensive hashing calculations efficiently. GPUs, originally designed for graphics rendering, happen to excel at these mathematical operations. This creates competition for the same chips that AI applications demand, compounding supply pressures we're already experiencing.Hardware Security Modules represent another specialized requirement. These devices store and protect cryptographic keys, requiring semiconductors built to strict security standards. Unlike general-purpose processors, HSMs must meet specific compliance requirements while handling cryptographic operations at scale. Every blockchain network that processes sensitive transactions needs this type of hardware infrastructure.Infineon Technologies offers a practical example with their NFC cryptochips that combine blockchain verification with physical products. You can attach these chips to items and scan them to authenticate origins, creating a bridge between digital blockchain records and physical supply chains. This isn't theoretical; it's addressing real counterfeiting problems with measurable hardware requirements.The technical demands go deeper than most realize. Blockchain networks require nodes that can process and verify transactions continuously, demanding consistent computational power rather than the burst processing typical of many applications. This sustained load creates different thermal and reliability requirements for the underlying semiconductors.The Foundry ResponseMajor manufacturers aren't waiting to see how this plays out. TSMC has implemented blockchain technology within its own supply chain operations, tracking components from raw materials through finished products. When the world's largest semiconductor foundry adopts blockchain for internal operations, that tells you something about both the technology's maturity and the hardware demands it creates.IBM and Samsung's collaboration on blockchain-based component tracking represents another significant development. They're building platforms that require substantial processing power to handle supply chain verification across multiple manufacturing partners. Each verification operation demands computational resources, multiplying hardware requirements as adoption scales.The industry has committed $185 billion in capital expenditures for 2025, expanding manufacturing capacity by 7%. While AI applications capture headlines, blockchain infrastructure contributes to demand pressures driving these investments. TSMC expects its AI server processor revenue to triple in 2024, reflecting how blockchain and AI applications often converge on similar hardware requirements.Manufacturing processes themselves are adapting to include blockchain verification layers. This creates additional computational overhead during production, requiring more processing power throughout the manufacturing workflow. It's a feedback loop where blockchain adoption drives chip demand while simultaneously being integrated into chip production processes.Securing the Chain That Builds the ChipsThe semiconductor industry faces a $75 billion problem with substandard components entering global markets. Blockchain applications are addressing this directly by tracking manufacturing equipment and components throughout production cycles. Manufacturing machines get registered on blockchain networks with unique identities, creating tamper-evident records of their performance and maintenance history.This approach reduces the manual burden of identifying counterfeit or substandard components. Instead of relying on post-production testing and documentation review, blockchain verification can flag issues during manufacturing. The computational requirements for maintaining these verification networks add another layer to semiconductor demand.Supply chain security applications demonstrate blockchain's practical value beyond financial transactions. When you can track a chip from raw silicon through final assembly with cryptographic certainty, you're solving real problems that cost the industry billions annually. This creates sustainable demand for blockchain infrastructure rather than speculative applications.Beyond the Hype CurveIndustry projections show semiconductor sales reaching $697 billion in 2025 - an 11% year-over-year increase - with expectations of hitting $1 trillion by 2030. Blockchain infrastructure represents a meaningful portion of this growth, though it's often overshadowed by AI applications in market analysis.The sustainability of blockchain's semiconductor demand depends on practical applications rather than speculative projects. Supply chain verification, secure authentication systems, and infrastructure applications that solve measurable problems will drive continued hardware requirements. Gaming platforms, financial networks, and enterprise verification systems all represent sustained demand sources.Manufacturing capacity constraints affect blockchain infrastructure deployment just like other applications. The 7% capacity expansion planned for 2025 addresses multiple growth drivers simultaneously, with blockchain contributing to demand pressures alongside AI, automotive electronics, and traditional computing applications.Current capital expenditure commitments total $185 billion across the industry, but this may not adequately address combined demand from AI and blockchain applications. Lead times for advanced semiconductor manufacturing equipment often exceed two years, meaning today's capacity decisions determine 2027's supply capabilities.The Semiconductor StakesWhile AI dominates technology headlines, blockchain infrastructure creates parallel demand streams that compound manufacturing pressure. 70% of executives expecting a significant blockchain impact aren't responding to speculation - they're planning for technical requirements they can measure and quantify.We're approaching an inflection point where blockchain applications prove their practical value while requiring substantial hardware infrastructure. Supply chain security, authentication systems, and verification networks all demand specialized semiconductors that didn't exist five years ago. The industry faces a dual challenge: scaling production capacity while implementing the very blockchain technologies that help secure that scaling process.The 2028 market projections provide a concrete checkpoint for these predictions. Whether blockchain infrastructure delivers on its $1.5 billion semiconductor market potential depends on the continued adoption of practical applications that solve real problems rather than speculative use cases.
Thursday 3 July 2025
SAC Group, member of WPG Holdings, partners with Axelera AI to expand into edge AI market and map out new blueprint for smart applications
As generative AI and edge computing technologies continue to evolve, Edge AI has become a key driving force for industries pursuing smart transformation. Recognizing this opportunity, Silicon Application Corp. (SAC) Group, a subsidiary of WPG Holdings and a global leader in semiconductor component distribution, announced today that it has obtained exclusive distribution rights for Axelera AI - an emerging European Edge AI chip company - in the Asia-Pacific region. The two parties will collaborate to combine SAC's strengths in supply chain integration with Axelera AI's technological innovation, aiming to help equipment manufacturers accelerate the development of Edge AI devices, enable independent software vendors (ISVs) to innovate more rapidly, and enhance the capabilities of system integrators (SIs) in deploying their smart applications."To promote the widespread adoption of AI, we launched the global partner program - the Axelera Partner Accelerator Network Program. We are pleased to have WPG Holdings' SAC Group as our core strategic partner in the Asia-Pacific region, working together to accelerate the implementation and application of Edge AI solutions," said John Wilkins, Global Channel Sales Director at Axelera AI. Wilkins pointed out that SAC brings deep technical resources, long-standing customer relationships, and a strong commitment to innovation - qualities that make them the ideal partner to bring our transformative AI technologies to broader markets. Through our partnership with SAC, we believe that equipment manufacturers, SIs, and ISVs will be able to adopt Axelera AI's highly flexible, high-performance, and scalable Edge AI solutions more quickly. This will shorten development cycles from proof of concept to mass production, and accelerate the deployment of smart applications from edge to cloud across various fields such as security surveillance, drones, retail, healthcare, and industrial automation.""The SAC Group's vision of promoting smart applications through Edge AI aligns perfectly with Axelera AI's core mission of driving the widespread adoption of AI technologies," said Lawrence Chang, Vice President of Tactical Development BU at WPG Holdings' SAC Group. He noted that many of the product lines currently distributed by SAC already support Edge AI applications, spanning diverse scenarios such as smart manufacturing, smart buildings, and in-vehicle systems. This collaboration will further enhance and complete SAC's Edge AI product portfolio. Looking ahead, SAC will continue to strengthen its technical support capabilities, offering customers end-to-end services - from early-stage evaluation and product selection recommendations to system integration testing. This will help clients adopt Edge AI technologies and solutions more efficiently, as they move together toward a fully intelligent AI-driven future."In the first half of 2025, WPG Holdings achieved the highest global revenue among semiconductor distributors and has consistently led the industry in technological investment related to computing chips," said Eddie Chou, Senior Project Director in the CEO Office at WPG Holdings. Since the rapid rise of AI in 2023, WPG Holdings has launched an AI acceleration initiative focused on three core areas. On the product front, the company has expanded its AI product portfolio to meet growing customer demand and has been recognized by over 30 suppliers of AI-related chips - including those specializing in compute, power management, and other critical components supporting AI applications. It is now fully capable of providing a complete range of hardware and resources required for AI servers and data centres. In terms of services, WPG Holdings has trained its sales teams to gain a deep understanding of AI project requirements. In addition to basic component support, it now offers advanced, integrated sales services for system-level servers and AI computing power management. From an investment perspective, WPG Holdings has participated in strategic investments in advanced-process IC design, helping more strategic partners build a thriving and collaborative AI ecosystem. Aspiring to become a key player in the AI industry chain, WPG Holdings' reliable partnership with Axelera AI serves as a concrete step toward realizing a mutually beneficial and prosperous future for the whole industry.
Wednesday 2 July 2025
AI meets RNA: InFocus's advanced RNA-targeted small molecule therapeutic platform accelerates drug development
While biopharmaceutical technologies have made substantial strides in treating diseases, drug research and development continue to face a multitude of obstacles, such as complex methodologies, high R&D costs, limited target options, and poor translational efficacy. This is particularly relevant for gene or RNA-level abnormalities, which are not effectively addressed by protein-targeting therapeutics, limiting their clinical impact.The rapid advancement in RNA biology has triggered a surge of interest and investment from pharmaceutical companies and the capital markets, accelerating the explosive growth of the RNA therapeutic sector. InFocus Therapeutics is leading this momentum, combining AI with RNA science to create the world's first and currently only "RNA-targeted small molecule drug generation platform." In just four months, InFocus successfully generated two series of drug candidates for spinal muscular atrophy (SMA) and advanced them to the animal efficacy and toxicity testing phase, demonstrating exceptional research and development efficiency.According to Emily Fang, CEO and founder of InFocus Therapeutics, the traditional drug development method is labor-intensive, time-consuming, and marked by a low success rate. In recent years, major pharmaceutical companies have poured significant resources into RNA-based therapies. Yet, a lack of innovative R&D strategies continues to limit their potential. On the other hand, most AI startups struggle to deliver true breakthroughs in this domain, largely due to a lack of real-world experience in early drug discovery and the complex process of transforming molecules into clinically viable, patentable drug candidates.InFocus rapidly gained traction by leveraging the team's deep expertise in AI and RNA technology, the discovery of novel chemical entities that differentiate from competitors and existing marketed drugs. In just four months, they identified two sets of promising therapeutic candidates, an achievement unmatched in the industry and now a key competitive advantage.40x Faster R&D Efficiency: Unlocking the Power of RNA-Targeted MedicinesFounded in 2023, InFocus Therapeutics specializes in the development of small-molecule chemical drugs that directly target RNA, an area poised to reshape modern therapeutics. The company brings together a multidisciplinary team of experts spanning molecular biology, AI algorithms, finance, and RNA drug discovery. Designed to tackle high-impact, hard-to-treat conditions such as neurodegenerative diseases and metastatic cancers, InFocus has built a proprietary RNA-targeted small molecule drug generation platform that integrates advanced AI algorithms and specialized databases to automate the lead discovery process, from molecular design and generative chemistry to structure prediction and experimental validation. The result: a breakthrough platform that delivers up to 40 times greater R&D efficiency, rapidly uncovering best-in-class drug candidates far beyond the reach of conventional methods.Using spinal muscular atrophy (SMA) as a case study, InFocus's platform efficiently identified two promising compounds, IFT-0000002 and IFT-0000003, within a remarkably short timeframe. Both compounds demonstrated strong biological activity, good oral absorption, and toxicological profiles, establishing a solid foundation for further development. CEO Emily Fang highlighted that the InFocus platform achieves a 40% success rate in RNA-targeted drug development, far surpassing the typical ~1% success rate of traditional small-molecule drug development. This high efficiency represents a core competitive advantage for the company.Raising Capital to Accelerate Pipeline ExpansionInFocus is currently raising a US$1 million angel round to support critical next steps. The funds will be deployed to complete the preclinical development and animal studies for the SMA program, and to launch R&D on two new RNA-targeted cancer programs for drug-resistant prostate cancer and colorectal cancer. These efforts will significantly expand the depth and breadth of the platform's therapeutic reach.Commitment to Establish and Grow Taiwan as a Clinical Hub for Strategic CollaborationsTaiwan boasts a comprehensive health insurance system, a world-class medical system, a highly digitized medical record system, and a plethora of clinical research capabilities. It has long been the favored site for clinical trials for international pharmaceutical corporations. InFocus Therapeutics views Taiwan as a key launchpad for clinical collaboration and regional expansion, and plans to set up a subsidiary in 2025. It will engage with the company's proposed clinical trial application framework for 2026 and formally commence the product's admission into the clinical validation phase, while also accelerating the recruitment of local talent and partnering with clinical collaborators.Emily shared that when seeking collaboration opportunities in Taiwan, she fortuitously connected with Taiwan Tech Arena (TTA) and joined its startup incubation program. InFocus gained valuable support and mentoring resources from the Mosaic accelerator, laying the groundwork for future investment and global market expansion. "Soon after our partnership with TTA, we engaged in the 2024 Founder Pitch event, which markedly improved our visibility and fostered connections within the investment community."Alongside its R&D milestones, InFocus is actively expanding its international investor network via TTA, aiming to accelerate the completion of its angel round fundraising and reinforce InFocus's leadership position in RNA-targeted small molecule therapeutics. With its AI-powered platform, high development success rate, and global footprint, InFocus is charting the path toward long-term growth in one of biotech's most promising frontiers.InFocus-BFMTV: InFocus founder Emily Fang was interviewed by renowned economist Nicolas Doze on BFM Business, France's number one financial news channel, and announced that it had received the Sanofi-sponsored 2024 Oncology "Golden Ticket" prize.InFocus Lab in a Loop: InFocus's proprietary artificial intelligence RNA drug discovery technology can generate high clinical potential therapeutic candidates targeting pathogenic RNA within a few months.
Tuesday 1 July 2025
Using AI to address chronic disease care challenges: AIM opens door to precision care for all
In an aging society, the care of chronic diseases has become a major social burden and has imposed significant impacts on the healthcare system. Elderly patients often face complicated medical procedures, unclear disease progression, and overcrowded settings in large hospitals. In 2010, Eric Lee returned to Taiwan from abroad to care for his mother, who had long suffered from chronic diseases related to bone and heart health, and went through firsthand the complexities of the healthcare system-ranging from resource bottlenecks to delayed diagnoses. Motivated by his personal journey, Lee co-founded AIM with physician Dr. Michael Yen in 2018. The company officially launched its product development activities in late 2020, focusing on solving pain points in chronic disease care for the elderly population through state-of-the-art AI diagnostic imaging technology.With Taiwan on the brink of becoming a super-aged society, diseases such as osteoporosis and degenerative arthritis are increasingly prevalent. However, early symptoms are often unnoticed or ignored, causing many patients to miss optimal treatment windows. Traditional diagnosis modalities like DXA bone densitometry or costly CT/MRI scans are time-consuming to schedule and not widely available, making large-scale community screening difficult. Furthermore, cardiac function assessment relies heavily on cardiologists operating advanced ultrasound systems, creating a bottleneck in primary care and public health services.Making Precision Care Accessible: The AI Revolution in X-ray and UltrasoundAccording to Lee, AIM's core technology leverages patented proprietary deep learning algorithms to analyze medical images from standard X-ray and ultrasound equipment. Its flagship solution, DeepXray, delivers automated bone density and joint degeneration analysis in under 30 seconds, generating clinically relevant reports. Validated by leading medical institutions including Taichung Veterans General Hospital, Taipei Medical University Shuang-Ho Hospital and Far Eastern Memorial Hospital, DeepXray achieves over 0.9 correlation with gold-standard DXA scans and maintains a world-class absolute error below 0.05 g/cm2.The company's second core product, DeepSono, focuses on automated cardiac ultrasound analysis. Using portable ultrasound devices, DeepSono autonomously outlines myocardial contours and calculates ejection fraction (EF) and global longitudinal strain (GLS)-traditionally requiring manual assessment by cardiologists.AIM's solutions have earned Class II medical device certification from authorities including the US FDA, Taiwan's TFDA, Japan's PMDA, and Vietnam's IMDA, and are expanding to other targeted countries-making it one of the few Taiwanese medical AI startups with comprehensive international certifications.Scalable Business Model and Global ExpansionTo accelerate adoption, AIM offers medical institutions a printer and cartridge business model, including a one-time software license fee at 10% of the full cost to lower entry barriers, followed by a SaaS-based monthly subscription or usage-based billing. Since its 2024 certification, AIM's solutions have been deployed in over 50 clinics and hospitals across Taiwan and Vietnam, with entry into the Japanese market underway."In a competitive global medical AI market, AIM's edge lies in delivering quantitative diagnostic data-not just risk classification," said Lee. While most medical AI solutions on the market only classify patients into high-, medium-, or low-risk categories, AIM goes beyond simply identifying risk levels and provides actionable and clinically relevant number-based metrics. AIM also prioritizes algorithm optimization and platform-based, fully automated diagnostic workflows, significantly enhancing clinical efficiency. Its long-term mission is to democratize precision care by pairing affordable devices with advanced AI technology so that the general public around the world will all be taken care of.Future Roadmap: Elderly Mobility and Cardiac CareLee stated that AIM will continue to expand its product portfolio centered around elderly mobility and cardiac care by leveraging widely available X-ray and ultrasound devices in primary care settings. Upcoming features include sarcopenia risk prediction, enhanced integration of handheld ultrasound with AI, and new developments in abdominal ultrasound, AI-assisted chest disease detection, and preoperative anesthesia risk analysis.Lee also underscored the evolution of Taiwan's startup ecosystem, which in recent years has matured significantly, especially in terms of global outreach. Taiwan Tech Arena (TTA) has actively facilitated overseas market connections and aligned investment resources, supporting AIM in its global expansion. In addition to tapping into international markets, Lee recommends that the government and hospitals provide more PoC/PoB habitats to accelerate the development and global reach of Taiwan's medical AI sector, increase clinical deployment opportunities for startups, and bring precision care to communities around the world.
Tuesday 1 July 2025
System cybersecurity and code storage flash technology
This article outlines the various code storage flash technologies used today and their contribution to platform cybersecurity and resiliency. It also outlines the minimal set of requirements to meet the demands of cyber protection in 2025 and beyond.Non-volatile storage media contain all the essential platform assets. The platform code, operational data, user data, and various states of the platform. For that reason, the content of the non-volatile storage is considered to be the most vulnerable to cyberattacks. Such attacks may try to extract user data (privacy), platform or network data (stability and resiliency) or modify the platform code as an extended attack vector for more complex attacks.Previously, standalone devices with no network connectivity were common, requiring attackers to have physical access to compromise system code. Such equipment had to be physically manipulated in order to gain access to the content of the non-volatile (NVM) storage. In those platforms, we would have one of three possible configurations for the non-volatile storage:*Embedded flash - where the NVM would be part of the controller silicon*External flash - where the NVM would reside side by side with the controller device*Multi-Chip-Module (MCM) - where the NVM die would be placed in the same plastic package as the controllerEach of these configurations would have its pros and cons. The embedded NVM would be the most secure, as it could be made almost impossible for an attacker to probe and access the embedded NVM portion of the silicon die. However, embedded flash does not allow easy expansion in storage capacity, and is not available in advanced manufacturing processesThe external flash would prove to be the easiest to attack. Connecting simple bus "sniffing" equipment to the flash would allow an attacker to extract information, modify code and perform any other actions required to compromise the platform and its data.Unauthorized NVM Access & Bus SniffingThe MCM approach seems to have the benefit of limited physical access to the NVM Flash device, while offering better cost structure. However, it is fairly easy today to decap (open) any chip package and access internal connecting signals, making the MCM approach nearly as (non) secure as the external flash configuration.MCMIn order to mitigate attacks on NVM, a special type of flash was developed, called "Secure Flash" where the interface signals between the controller and flash are encrypted. This prevents the traditional probing attacks from being successful.Secure flashWith modern devices increasingly connected to networks, security risks have expanded beyond physical threats to remote cyberattacks. All the above attacks required physical access to the attacked platform because of no networking connectivity. This rather limited the scope of such attacks. However, most of the platforms today, regardless of their functions, become network-connected. This trend is driven by a few key concepts:*Extended functionality (by using network resources)*Ease of use*Remote management and information/statistics collection*Frequent, automatic firmware updates*Easier deployment and life cycle controlThe key driver for network connectivity is remote firmware update. This is mandated by all the latest standards and regulations, such as EU RED EN18031, US CNSA2.0, and the coming EU Cyber Resiliency Act. It is well understood today that orderly firmware updates are a must in order to keep systems resilient and protected, and their users safe.Making platforms network-connected and set up for automatic remote system updates opens up a whole new landscape of potential security attacks. This makes the physical protection, mentioned here before, an outdated issue in most cases, with the exception being open-loop financial transactions and ID cards (such as passports).Modern platforms must be protected from a completely different set of threats:*Unprotected and outdated updates - attempt to force a platform to revert to an older version of its software with known vulnerabilities or lack critical security features, making them easier for attackers to exploit.*Supply chain attacks - attempt to insert Trojan hardware or install malicious software at some point within the manufacturing, transportation, or distribution of platforms.*Storage breeching (keys, user data, credentials theft) - unauthorized access or compromise of stored sensitive information.*Threats to the resiliency of the platforms - attempt to infect systems with malicious code, exploit vulnerabilities in platform firmware, or make the platform unavailable by destroying critical data in NVM.It is well understood that large-scale attacks can be leased upon millions of devices without a need to physically access any of them. It is thus required to implement protection mechanisms against such potential attacks.Firmware update protection - latest standards call for protection of firmware updates in terms of integrity, authenticity, and freshness, allowing only a complete, signed, and newer version of firmware to replace an existing version. Moreover, the cryptographic signature mechanisms to be used must be quantum-safe, meaning algorithms developed to resist being attacked by quantum computers.Supply chain protection - this topic is at the top of the list for potential large-scale attacks published by EU ENISA in 2024. Due to the distributed nature of supply chains used in the production of electronic equipment, it is relatively easy for rogue players to replace key components such as the NVM devices and basic firmware code with devices and code that include malicious capabilities. These capabilities can then be employed to unleash a large-scale attack on infrastructure.Storage protection - if a platform is allowed to access any information in the NVM at any time, a rogue piece of code, even temporarily loaded into memory, can scan the NVM and access secret or sensitive data, transmit it back to the attacker, and even modify this data. It is essential to maintain privileged access restrictions on such data in the non-volatile memory.Platform resiliency - as more and more platforms become online and network-connected, it is essential to ensure they are kept in correct working order or otherwise gracefully brought down to ensure no harm is done if they are somehow compromised. To facilitate this, strong internal diagnostics of the NVM content (code and data) are required, and in the case of some malfunction, allow the platform to either fix the problem or go offline and shut itself down. Platform firmware resiliency is detailed in NIST SP 800-193.Given these contemporary threats, it is clear that NVM packaging technology plays a diminished role, if any, in platform cyber protection. The focus should be on using the appropriate NVM built-in security capabilities and configurations to minimize and eliminate these threats.The Winbond W77Q Secure Flash is introduced as a robust solution to address the outlined threats. The key features of W77Q Secure Flash relevant to the document's topics include:*Code and Data Protection: robust protection for both code and data, making it exceedingly difficult for hackers to tamper. RoT implementation follows the TCG DICE attestation mechanism.*Authentication: Winbond Secure Flash devices employ stringent authentication protocols, ensuring that only authorized actors and software layers gain access. *Secure Software Updates with Rollback Protection: The devices facilitate remote secure software updates while safeguarding against rollback attacks, ensuring that only legitimate updates are executed. To maintain the highest level of security and integrity during software updates, W77Q employs Quantum-Safe Leighton-Micali Signature (LMS) algorithms, as recommended by NIST Special Publication 800-208. This method guarantees the authenticity and integrity of the updated software, thereby providing an additional layer of security for years to come.*Platform Resiliency: following NIST 800-193 recommendations, unauthorized code changes are automatically detected, enabling the system to automatically recover to a secure state and disturbing potential cyber threats.*Secure Supply Chain: The origin and integrity of flash content are guaranteed by Secure Flash at every stage of the supply chain. W77Q implements remote attestation based on LMS-OTS (NIST 800-208). This advanced approach effectively prevents content tampering and misconfiguration during platform assembly, transportation, and configuration, safeguarding against cyber adversaries.Winbond W77Q Secure Flash addresses critical security concerns discussed in the paper, providing strong protection for code and data storage while ensuring platform integrity, resiliency, and compliance with emerging cybersecurity frameworks.For more information on how Winbond can support your security and compliance needs, visit Winbond's website or contact Winbond directly, or download the latest Hardware Security White Paper.
Tuesday 1 July 2025
Bytelingo's professional services redefine AI translation, attracting engagement from publicly traded companies
Despite the increasing prevalence of AI applications, the majority of free translation solutions are incapable of satisfying the specific translation needs of enterprises in professional sectors, as they are designed to address generic situational requirements. Organizations that depend on free AI translation frequently must allocate substantial human resources to review and proofread, a process that is both labor-intensive and time-consuming. This holds true when using standard financial statements or product manuals as examples.Bytelingo has acknowledged this challenge and has implemented generative AI to offer translation solutions that are exceedingly precise, particularly for professional publications like annual reports and ESG reports. As of now, Bytelingo's services have been adopted by more than 150 publicly traded companies, with a renewal rate of 90%, indicating a significant level of market confidence.Ringo Chang, founder and CEO of Bytelingo, observed that common AI translation services, including ChatGPT, often produce "highest probability" assumptions based on model algorithms when faced with numerous proper nouns and terminologies (such as full names and abbreviations of products, subsidiaries, affiliated entities, government agencies, etc.), resulting in a considerable incidence of translation errors.Conversely, Bytelingo utilizes a different technique. It initially collects and classifies clients' historical translation papers and official website data, then integrates this information with existing professional databases, including accounting subjects and global standards, to improve translation accuracy and consistency.Exclusive technology outperforms peers with annual report translation taking only 3 hoursChang, with considerable expertise in AI algorithms, acknowledged the Financial Supervisory Commission's mandate for listed companies to submit English financial statements and ESG reports, resulting in the establishment of Bytelingo in June 2022 to tackle the challenges of professional translation for corporations.Initially, Bytelingo relied on proprietary AI models. At the end of 2022, as generative AI and large language model technologies progressed rapidly, the firm modified its product development strategy, concentrating on the development of an AI translation platform based on efficient large language models accessible in the market.Currently, Bytelingo maintains comprehensive databases comprising over 30,000 entries in accounting and sustainability, along with proprietary optimization methods that provide swift access to clients' company-specific information. After completing data preparation, the capabilities of generative AI are effectively governed through its advanced RAG (Retrieval-Augmented Generation) technique, yielding more accurate and consistent translation outputs."Our service requires only one morning to finalize the data configuration, followed by merely three hours to translate the entire annual report," Chang emphasized that this efficiency substantially surpasses that of conventional manual translation, which frequently requires a month. Moreover, Bytelingo's pricing is only 50% to 70% of that of traditional translation agencies, enabling faster customer acquisition. He claimed that Bytelingo's publicly traded clients are expected to surpass 200 by the end of 2025.Full TTA support boosts global prominenceSince its inception, Bytelingo has garnered multiple honors and been selected for various startup awards and government subsidies. This includes the Taipei City Government's 2023 SITI Star investment-potential start-up team, the Ministry of Digital Affairs' Go Next Digital review award, and the endorsement from the Small and Medium Enterprise and Startup Administration of Taiwan's Ministry of Economic Affairs for entry into TAcc+.Upon becoming a TTA (Taiwan Tech Arena) team, Bytelingo received mentorship and support from the STARTUP101 accelerator, which facilitated the company's business expansion, fundraising, brand visibility, and other critical areas.Chang stated that TTA resources and initiatives have afforded Bytelingo a global perspective. During a visit by a large German media conglomerate to TTA, Bytelingo was granted the opportunity to present on stage, therefore enhancing the company's brand recognition and providing substantial benefits in following endeavors to secure venture capital funding. In the future, Bytelingo will strengthen its collaboration with TTA and accelerate its global expansion.Bytelingo is not only translating annual and ESG reports but also diversifying its applications, encompassing product catalog and technical document translation, AI conference support, and AI product introductions, to more effectively meet the complex needs of companies in cross-linguistic communications.Considering international market prospects and the mandate from Japan's financial regulatory bodies for local companies in Japan to provide English annual reports, Bytelingo will utilize its successful model from Taiwan to enter the Japanese market and exploit opportunities in the AI professional translation service industry.Bytelingo offers a variety of AI platform modules to satisfy the unique translation process requirements of businesses. Credit: Company
Friday 27 June 2025
Hyundai Exter: Why this micro-SUV perfect choice for city driving
City driving requires a car that's small, agile, versatile, and loaded with the basics to simplify the daily grind. Meet the Hyundai Exter - a micro-SUV that is prompting some serious hype in all the right ways! Targeted at urban drivers, the Exter is a versatile, safe and tech-rich offering that seems perfectly suited to city dwellers.Perfectly Sized for Urban StreetsLet's start with its size. The Hyundai Exter is small enough to weave through traffic and fit into tight parking spots, yet spacious enough to keep occupants comfortable during long commutes.Hyundai Exter specificationsDimensionMeasurement (mm)Overall Length3815Overall Width1710Overall Height1631Wheelbase2450Source: CompanyThese dimensions make it an ideal city car - compact enough to squeeze through narrow lanes, yet with a generous wheelbase that offers good cabin space and stability.Engine Options That Suit Every Urban NeedOne of Exter's strongest suits is its range of powertrain options. Whether you're looking for traditional petrol or a greener, cost-saving CNG variant, Hyundai gives you four practical choices:Hyundai Exter engine specificationsEngine TypeFuel TypeTransmission1.2 l Kappa Petrol MTPetrol5-speed Manual1.2 l Kappa Petrol AMTPetrolSmart Auto AMT1.2 l Bi-fuel Kappa Petrol with Hy-CNG MTPetrol + CNG5-speed ManualSource: CompanyThe automatic (AMT) variant is a blessing in city traffic, where stop-and-go situations can make manual driving a chore. Meanwhile, the bi-fuel CNG options cater to the cost-conscious without compromising performance.Built for City ComfortWhile the Exter is modest in size, it doesn't skimp on comfort. From entry-level models to higher trims, it delivers essential urban creature comforts:*Smart Electric Sunroof: A welcome addition that makes city driving more enjoyable, especially during winter or in scenic neighbourhoods.*Push Button Start/Stop: No more fumbling with keys - just step in and go.*Wireless Phone Charging: Keeps your smartphone topped up without messy cables.*Power Windows: Available for all four doors, adding convenience for passengers, too.*8-inch HD Infotainment System: With Apple CarPlay and Android Auto, staying connected on the move has never been easier.All of this is wrapped in a smartly designed cabin with thoughtful features like a passenger vanity mirror, front fast USB-C charger, and steering-mounted controls for audio and Bluetooth.Safety First, AlwaysCity driving isn't without its hazards - tight junctions, busy crossroads, and unpredictable traffic demand a vehicle that's as focused on safety as it is on style. The Hyundai Exter doesn't disappoint:*6 Airbags: Driver, passenger, side, and curtain airbags ensure full occupant protection.*ABS with EBD: Keeps the car stable under sudden braking.*Electronic Stability Control (ESC): Great for slippery or uneven roads within city limits.*Speed Sensing Auto Door Lock: Ensures added safety once the car gets moving.*3-point seat belts for all seats: Including rear passengers - a rare attention to detail in this segment.These features are often found in premium SUVs, yet Hyundai has brought them to the micro-SUV category.Designed to Be SeenJust because you're driving a compact vehicle doesn't mean you should go unnoticed. The Exter's bold styling helps it stand out in city traffic:*Bi-function Projector Headlamps: Offer better visibility at night.*LED Daytime Running Lamps (DRLs) and LED Taillamps: Enhance visibility while adding a stylish signature.*LED Turn Indicators on Outside Mirrors: A detail you'll appreciate during city manoeuvres.*Black-Out A, B Pillars and Window Line: Creates a seamless, sporty look.*Front & Rear Mudguards: Protect the body from everyday splash and grime.The overall design feels mature, sporty, and modern - ideal for young professionals, couples, and even small families.Suspension and Ride QualityCity roads can be anything but smooth, and this is where Hyundai's suspension tuning shines:Hyundai Exter suspension specificationsSuspension TypeSetupFront SuspensionMcPherson StrutRear SuspensionCoupled Torsion Beam AxleSource: CompanyTogether, they provide a comfortable ride even over potholes and poorly surfaced roads - exactly what city dwellers need.Small Car, Big TechHyundai's Bluelink connectivity suite injects an extra dose of smart convenience into the Exter, offering over-the-air (OTA) updates for maps and infotainment. Location tracking, remote engine start (in case of select models), and vehicle health monitoring can all be accessed from your phone.Fuel Efficiency and PracticalityExter holds 37 litres of juice, so you won't need to refuel that often. CNG versions provide great savings for those who do a lot of driving. Whether doing school runs, office hops, or weekend errands, this micro-SUV is engineered to make your life simpler.Final ThoughtsIn a city environment where convenience, safety, efficiency, and modern features matter the most, the Hyundai Exter checks all the right boxes. It's got the footprint of a hatchback but the road presence and features of a much larger SUV. From its peppy yet frugal engines to its sunroof, six airbags, and connected tech, this is a car that's genuinely made for the rhythm and chaos of urban India.
Thursday 26 June 2025
Graser TECHTALKS 2025: AI enables intelligent integration in electronic design
Artificial intelligence (AI) is transforming the electronic system design methodology. In early June, Graser hosted its annual flagship user event, Graser TECHTALKS 2025. Under the theme "Smart Tri-Fusion: AI Connected Design-Analysis-Manufacturing," Graser brought together experts from upstream and downstream industry partners, as well as internal specialists, to discuss the use of electronic design automation (EDA) tools alongside AI to boost design efficiency and quality. They provided insights on market trends and practical case studies to hundreds of attendees.Lillian Pan, General Manager of Graser, underscored at the conference's inauguration that the company has been engaged in the electronics sector for over three decades, dedicated to collaborating with partners to deliver innovative and cost-effective design solutions. Graser is proud to be one of Cadence Design Systems' most valued channel partners, having earned multiple Cadence Channel Partner of the Year awards over the years. In the future, Graser will continue to serve as a "technology connection" and lead the industry into a new era of intelligent design.Cadence presents intelligent PCB design frameworkMichael Jackson, corporate VP and GM of the system design and analysis group at Cadence, delivered the opening keynote of the conference. He discussed the use of advanced tools, AI, In-Design Analysis (IDA), and GPU-accelerated computational assistance to construct superior electronic products at a faster pace.Jackson stated that Cadence offers a three-layer electronic system design and analysis platform that is extensively utilized in PCB design, system packaging, and multi-physics simulation analysis. The system design layer, which is represented by tools such as OrCAD and Allegro, is at the top of the platform. The system analysis layer, which is at the bottom, utilizes multi-physics simulation analysis tools such as Optimality, Voltus, Clarity, Sigrity, Celsius, and others to address a variety of intricate issues in electronic and electromechanical systems.In-Design Analysis, which is situated between design and analysis, enables designers to initiate simulation analysis tools directly in the design interface and provide immediate feedback on the results, thereby enhancing design efficiency and accuracy.By integrating three distinct AI technologies into this platform, Cadence offers more efficient and comprehensive solutions for the design, analysis, and development of electronic systems. Optimization AI prioritizes cost-effectiveness, performance, and efficiency in order to identify the optimal solution. Jackson noted that Cadence's X AI platform significantly streamlines the PCB layout process by utilizing optimization AI. In fact, X AI has the ability to reduce the time required to complete a circuit board layout assignment that had previously required several days to be completed manually to a mere few minutes.Generative AI allows designers to engage with design tools in natural language, generate Skill codes, and query design parameters, thereby reducing entry barriers and increasing productivity. Jackson described a future scenario in which designers can issue high-level goals, such as cost constraints and lifespan requirements, and Agentic AI will automatically search for feasible solutions, optimize designs, and even produce multiple layout versions to achieve truly autonomous design in the context of agentic AI applications.Jackson specifically referenced the Cadence Millennium M2000, a new generation supercomputer featuring NVIDIA Blackwell systems, in relation to GPU acceleration. This platform has the potential to significantly enhance the efficiency and quality of the electronic design process by increasing simulation efficiency by dozens of times.Robust computing powers electronic design, accelerating Edge AI applications and AICharlie Shih, Product Engineering Group Director at Cadence, delivered additional keynote presentations at Graser TECHTALKS. His address was titled "High Computing Power Drives New Changes in Electronic Design and Optimization." He detailed how Cadence's Clarity 3D Solver integrates with tools like Allegro and Microwave Office to enable In-Design Analysis, significantly shorten development cycles, and assist design teams in accurately simulating complex structures such as packages, PCBs, and antennas by combining advanced algorithms with massively parallel computing.Furthermore, Magic Pao, Associate VP at Advantech, a prominent industrial computer maker, shared insights on the evolution and future of edge AI applications. Magic indicated that the expense of AI inference has decreased markedly by 99.7% over the past two years. Coupled with improved computing platform performance and increased privacy protection requirements, 2025 will be a critical starting point for the explosion of Edge AI applications.Advantech focuses on five core areas: smart manufacturing, cities, transportation, and healthcare. It has extensively sketched out relevant solutions, including hardware modules, computing platforms, simulation and middleware software, and plans to deliver a higher computing platform to handle more advanced AI inference and robot control applications.Ansforce CEO Jefferey Chiu delivered a speech titled "Generative Artificial Intelligence Is Transforming New Industrial Trends". He explained the technical principles, application scenarios, and development limits of generative AI to the audience in simple and amusing terms. He also examined the impact and possible commercial prospects brought about by related new applications in other industries, which provided a fitting conclusion to the morning session of Graser TECHTALKS.Design and simulation engines work together to speed up developmentThe Graser TECHTALKS afternoon session provided a comprehensive demonstration of Graser's deep integration and intelligent upgrade solutions from two perspectives: design to manufacturing, and software platform to simulation verification. It also extended an invitation to industry partners to share successful application cases."Intelligent PCB Design" explored essential tools and application methodologies, including the key features of the recent Allegro/OrCAD X 24.1 upgrade and how Allegro X System Capture integrates circuit design and layout review to improve overall development efficiency. Graser also unveiled the GIMS system, which includes parts auditing, automatic parts selection, version control, and BOM management to assist the team in streamlining complexity.Additionally, Graser underscored the efficient DFM process of CAMPro and the proprietary techniques of GraserWARE in collaboration with Allegro to expedite PCB design. Eastek Technology demonstrated its ezPlan tool, which optimizes the PCB Panelization and Stackup Selection for Manufacturing flow, and Graser extended an invitation to MSI to clarify the use of the Graser New CM Import tool to simplify the layout process.The "Multi-physics Simulation Analysis" examined the signal and power integrity challenges in high-speed design, demonstrating the efficacy of the simulation technology and its implementation. The Graser team initially conducted a functional analysis of the Sigrity 2024.1 Hotfix and specifically presented the Graser Sigrity Scheduler and Graser Stackup Exchanger within its premier proprietary solution, GraserWARE MSAPack.The Graser Sigrity Scheduler is capable of autonomously scheduling and executing Sigrity's SI/PI simulation tasks, while the Graser Stackup Exchanger can synchronize layered structure information across platforms to ensure that design data is highly consistent across multiple departments and tools.MPS was invited to demonstrate how to use the network analyzer with OptimizePI for PDN impedance simulation and experimental verification, ensuring the voltage stability of the AI chip. AUO presented its recent advancements in optical interconnect packaging design and reviewed the outcomes of its partnership with Cadence to enhance the optical communication driver design procedure. Compal Computer unveiled its internally developed Zero Touch simulation technology, significantly enhancing R&D efficiency and traceability via a fully automated process.Ultimately, Cadence showcased the Celsius Studio multi-physics integrated thermal simulation platform, highlighting the effective mastery of thermal management through actual applications. Graser effectively showcased the successful outcomes of synergizing partner resources to drive intelligent integration across electronic design ecosystem through case sharing and technical exchanges.The recently held 2025 Graser TECHTALKS, themed Smart Tri-Fusion: Smart Connected Design-Analysis-Manufacturing, highlights the use of EDA tools and AI to boost design efficiency and quality. Credit: Graser