ReadingResearchRadarInvestment framework
Sign in / Sign up
AllTechnologyProductsInvestingPublic marketsOther

Reading

2026-11-2034 posts

Refresh

US House Panel Reveals Commerce Dept Approved Over 879,000 AI Chip Exports to China, Including Military-Linked Firms

Structural
2026-11-20

Documents released by the US House Foreign Affairs Committee on November 19, 2026, reveal that the Department of Commerce's Bureau of Industry and Security (BIS) had approved the sale of over 879,000 artificial intelligence (AI) chips to Chinese companies as of May 2026. The approved buyers include tech giants such as Alibaba and Tencent, with some entities listed by the US Department of Defense as "Chinese military enterprises." This disclosure comes amid heightened scrutiny following the resignation of former Nvidia CEO Jensen Huang due to personal scandals.

The committee noted that between December 2025 and May 2026, applications for selling Nvidia H200 and equivalent AMD chips to China totaled 3.4 million. Applicants also included ByteDance, Tesla Shanghai, and Baidu. Committee Chairman Brian Mast specifically named Maginfra and New Matrix as suspected shell companies facilitating chip smuggling, citing a New York Times report that their offices were empty and linked to sanctioned entities.

Democratic Ranking Member Gregory Meeks criticized BIS for failing to enforce regulations, emphasizing the public's right to know how these decisions were made. The release relies on the Export Control Reform Act, which allows Congress to declassify export data if secrecy is deemed against national interest. While Commerce, Nvidia, and AMD did not immediately respond to requests for comment, the move intensifies doubts about the effectiveness of current control policies.

With Jensen Huang scheduled for closed-door questioning on December 1 and Commerce Secretary Howard Lutnick facing criticism for inaction, bipartisan lawmakers are pushing legislation to tighten restrictions on advanced AI chip sales to adversary nations. Additionally, AMD shares fell after short-seller Hunterbrook Media alleged it provided new firmware to blacklisted Chinese server maker IEIT Systems, highlighting the sector's geopolitical compliance vulnerabilities.

Sources:INSIDE

Research

Microsoft and Intel Deepen Partnership: First Use of EMIB Advanced Packaging in Custom Azure Chips, Securing Multi-Year Xeon Supply

Material
2026-11-20

U.S. tech giants Microsoft and Intel announced on November 20, 2026, an expansion of their decades-long partnership to advance Azure infrastructure development. The core breakthrough lies in Microsoft's first-time announcement of adopting Intel Foundry's EMIB (Embedded Multi-die Interconnect Bridge) advanced packaging technology for its custom Azure chips. This move marks further recognition by Microsoft of Intel's manufacturing capabilities within its AI infrastructure supply chain.

Beyond packaging technology, the two companies have established a multi-year capacity supply agreement for Intel Xeon processors, including the next-generation server CPUs codenamed Diamond Rapids. This long-term deal aims to ensure stable and predictable computing resources for Microsoft as it faces growing cloud and AI workloads. By combining Intel's chip design and packaging expertise with Microsoft's cloud engineering know-how, the partners seek to create differentiated infrastructure optimized for performance, efficiency, scalability, and security.

On the technical application front, Microsoft will collaborate with Intel on CXL-enabled Xeon processors to enhance Azure's memory efficiency. For instance, Azure M-series v4 virtual machines equipped with CXL memory and Intel Flat Memory Mode are reported to deliver up to 2.3x performance gains for high-demand workloads. Additionally, SQL backup speeds on Azure are expected to improve by up to 2.5x through Intel QuickAssist Technology (QAT). These system-level optimizations demonstrate deep integration in hardware co-design.

Sources:BenchLife.info

Research

Korean Web Builder I'mweb Reports AI Agent Submitted Over 100 PRs, Human Support Intervention Dropped Below 2%

Quick take
2026-11-20 08:56 GMT+8

Korean website building platform I'mweb disclosed recent advancements in its internal AI applications at its 'Techground 26' technical meetup. According to CTO Kim Hyung-seop, the team introduced AI agents capable of autonomously participating in work flows. One agent, designed as an 'AI colleague,' communicated with over 60 team members and independently wrote and submitted more than 100 Pull Requests (PRs), completing the entire process from code modification to deployment.

In the customer service sector, I'mweb achieved a higher degree of automation. Product engineer Lee Jung-min stated that by gradually optimizing the AI's judgment process for customer support responses, the rate of human intervention dropped from an initial 100% to less than 2%, eventually eliminating human judgment entirely in the live operational environment. This change relied on continuous validation and rule adjustments regarding the AI's accuracy.

Platform engineer Choi Sung-guk noted that for AI agents handling tasks, the key lies in setting criteria for when to ask humans for help and which decisions require joint human-AI judgment, rather than simply pursuing automated execution. Additionally, the infrastructure team emphasized strict pre-execution verification mechanisms when using AI for operations to ensure system stability. These cases demonstrate specific practices and quantified results of mid-sized internet companies integrating AI into core engineering and operational processes.

Sources:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Research

Korean AI Infrastructure Firm MOREH Unveils TIDE Framework at SC26, Claiming 1.66x Throughput Gain

Quick take
2026-11-20 07:43 GMT+8

South Korean AI infrastructure software company MOREH unveiled TIDE, an inference framework designed to optimize "speculative decoding," at the Supercomputing 2026 (SC26) conference held in Chicago. Speculative decoding accelerates Large Language Model (LLM) generation by having a smaller draft model predict multiple tokens which a larger target model then verifies. However, traditional implementations often suffer from performance degradation when user request patterns shift, as the statically trained draft model's predictions become inaccurate.

TIDE addresses this limitation by leveraging the "intermediate hidden states" generated by the large model during inference to retrain the draft model in real-time. This approach eliminates the need for separate data generation passes, reducing computational overhead while allowing the draft model to adapt to changing traffic patterns. The system monitors prediction accuracy and effectiveness continuously, pausing speculative decoding when benefits diminish and triggering retraining when patterns change.

According to results presented by MOREH at SC26, TIDE achieved up to 1.66 times higher inference throughput compared to baseline methods without speculative decoding. Additionally, the training time for the draft model was reduced by up to 3.02 times compared to existing online learning approaches. The company also demonstrated a heterogeneous computing strategy, running the large model's inference on Nvidia H100 GPUs while training the draft model on AMD Instinct MI250 GPUs, which reportedly improved overall system throughput by up to 1.22 times compared to using a single GPU type for all tasks. These figures are self-reported vendor measurements and have not been independently reproduced.

Sources:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Research

Arm China Launches 'Xingchen 100e' AIoT Prototype Platform with New Cortex-M42 Processor

Material
2026-11-19 12:02 GMT+8

On November 18, Arm China held a product briefing to provide an in-depth analysis of the Arm Cortex-M42 CPU and officially launched the 'Xingchen 100e' AIoT prototype platform based on this processor. Co-developed by Arm China and Arm, the Cortex-M42 is named 'Xingchen' STAR-MC3 in the Chinese market and has been formally integrated into Arm's global Cortex-M product roadmap.

Targeting smart endpoint devices sensitive to cost and area, the Cortex-M42 is built on the Armv8.1-M architecture and features three core capabilities: Arm Helium technology accelerates vector processing for stronger machine learning and digital signal processing performance; Arm Custom Instructions allow optimization for specific workloads; and a compact AHB-based design uses a mature microcontroller system architecture without complex memory subsystems. Compared to the predecessor Cortex-M33, its vector computing performance is tripled, and vector energy efficiency is doubled.

Regarding silicon area and power consumption, official data shows that in a configuration using TSMC's 22ULL process at 150MHz, the Cortex-M42 reduces silicon area by 10% and power consumption by 15% compared to the Cortex-M52, while delivering similar scalar and vector performance. This makes it suitable for designs that do not require rich memory subsystems.

The 'Xingchen 100e' is a heterogeneous computing platform combining the Cortex-M42 CPU with the Arm Ethos-U55 NPU. Arm China has built the system on an FPGA platform and successfully run typical AI use cases such as keyword spotting, object detection, and automatic speech recognition. Measured data indicates that compared to using the Cortex-M42 alone, pairing it with the Ethos-U55 improves keyword spotting inference performance by 95 times, object detection by over 132 times, and automatic speech recognition by over 340 times.

Zhu Xiaoming, CPU Product Director at Arm China, stated that the platform aims to provide various types of efficient computing power for low-power edge devices. However, he also emphasized that chip design is not simply 'building blocks'; merely combining IPs does not automatically yield competitive chips, and design companies still need to conduct extensive original architectural design and system optimization based on their understanding of applications and algorithms.

Sources:量子位

Research

Three Retired V100 GPUs Outpace $5,399 Mac Studio in Long-Context Inference for $1,042

Quick take
2026-11-19

According to Tom's Hardware, a homelab builder known as esologic benchmarked three used Nvidia Tesla V100 GPUs, finding that the ~$1,042 setup outperformed a $5,399 Apple M5 Max Mac Studio in inference speed for the Qwen 3.6 35B mixture-of-experts (MoE) model at long context lengths.

The tests, conducted using the llama.cpp framework, compared generation speeds across different quantization levels and context sizes. With 8-bit quantization (Q8), the three V100s achieved 100.8 tokens per second at 4K context, dropping to 72.7 tokens per second at 128K. The M5 Max Mac Studio, equipped with a 40-core GPU and 128GB of unified memory, was slightly faster at 4K (105.5 tokens per second) but fell behind at 64K and 128K, where the V100 cluster was approximately 19% faster at the longest context tested.

The V100s represent older 'e-waste' enterprise GPUs, whose advantage lies in combining VRAM capacity and bandwidth through multi-GPU parallelism, enabling larger models or longer contexts. While modern consumer devices like the Mac Studio may excel in short-context performance and energy efficiency, they were surpassed by the low-cost legacy hardware in this specific long-text workload. The builder also noted that the V100s consumed about half the energy of an equivalent number of P100s. Due to the use of different runtimes (llama.cpp vs. MLX), these figures should be taken as indicative rather than independently verified.

Sources:Tom's Hardware

Research

AWS Engineer Chambers: A Single Forward-Pass Decision Model Allows Agents to Skip Unnecessary Memory Retrieval

Quick take
2026-11-20 09:37 GMT+8

Say “good morning” to an AI agent, and it might search its long-term memory before saying hello back. Somewhere in that system, a service performs a semantic lookup to determine whether your greeting requires understanding your beverage preferences. Even if you take your coffee very seriously, this seems like a lot of infrastructure for two words.

Mike Chambers, a Senior AI DevEx Engineer at Amazon Web Services (AWS), presented a solution to The New Stack Editor-in-Chief Nick Lucchesi at AGNTCon + MCPCon North America 2026: using a decider model to handle the small judgments that accumulate inside an agent. These models are also known as decision models or System One models; they select from a fixed set of options or rate inputs, returning calibrated probabilities without generating free-form text. This concept was popularized by Jev of TypeSafe AI, and AWS’s Strands Labs released the open-source Strands Decider 2B, a model with two billion parameters, on October 1.

Chambers describes this approach as having an agentic-powered, language-aware if-statement within your code. Developers have used classifiers for years, but what distinguishes this generation of decider models is that general-purpose pretraining allows you to try them on new classification tasks without first training your own model. You provide natural-language context, ask a question, and receive a numerical score that the application uses to choose a branch. In the models Chambers describes, this score is returned in a single forward pass, with no token sequence requiring generation.

This makes decider models suitable for placement on the critical path of every request. Chambers experimented with the open-source Strands framework and Amazon Bedrock AgentCore Memory: he placed Strands Decider before the memory lookup to judge whether an incoming message required fetching information from past conversations. Greetings could pass straight through; requests dependent on what the user said earlier would still trigger a search. He also used a decider after retrieval: while a memory service can return the most relevant matches, some may be unrelated to the current task, so the decider can score each result for relevance before the agent incorporates it into the context.

One decision saves a service call, and another reduces the amount of irrelevant material the model must process. A model router can also use scores to send routine prompts to smaller models and reserve more capable models for complex tasks, saving resources from both the routing decision itself and the downstream work it avoids.

However, Chambers explicitly characterizes his memory experiment as experimental. While a number in a predictable JSON field makes it easy to integrate a decider into an application, determining whether that number is reliable enough to act upon requires testing.

"Will the number be correct? That’s for your testing to figure out," Chambers said. When stakes rise—such as when an agent is about to spend money rather than recommend a coffee—Chambers advocates for setting up checks outside the agent’s loop and requiring human approval for decisions that are difficult to reverse.

Sources:The New Stack

Research

CodeRabbit Tests GPT-6.2 Astra: Captures Seven More Known Issues but Precision Drops to 12%

Quick take
2026-11-20 04:00 GMT+8

In code review, capturing more issues does not automatically translate to a better review experience. After testing GPT-6.2 Astra within its own pipeline, AI code review company CodeRabbit found that while the new model did identify a few additional known issues, the cost was a significant increase in comment volume and a sharp decline in precision—a combination precisely what review tools aim to avoid.

CodeRabbit compared Astra 6.2 against the previous generation Astra integration using a benchmark set of 83 known issues. Astra 6.2 caught 58 issues (69.9%), an increase of seven over the previous generation's 51 (61.5%). However, reported comments rose from 331 to 847, and precision fell from 21.74% to 12.22%—meaning only about one actionable comment per eight published corresponds to a known issue in the benchmark. The authors emphasize that these results do not establish a pure model improvement, as the baseline configuration utilized Sol 6.1 and other models, with each configuration run only once.

The problem with comment inflation is not merely numerical. Of the 589 standard actionable comments, only 72 matched known issues; the remaining 517 may contain genuine problems not annotated in the benchmark, but the authors explicitly state that understanding their value requires reading them line by line. CodeRabbit notes that every comment demands the author distinguish useful findings from noise. When a pull request is filled with low-value feedback, authors may eventually stop reading or addressing comments, causing useful findings to be ignored as well.

Sol 6.2 exhibited a similar but less pronounced pattern. In tests involving 81 known issues, Sol 6.2's high-reasoning configuration caught two more issues than 6.1 (74.1% versus 71.6%), but comment volume increased from 187 to 290, and precision dropped from 35.3% to 25.9%, with recorded costs rising by approximately 48%. Again, these runs cannot isolate the effect of reasoning levels, as different models were swapped into different parts of the pipeline.

Regarding code generation, the authors did not run matched benchmarks but shared practical coding experiences. Astra often produced a useful initial interface version from short prompts and, in long tasks, could return to the original goal after deviations. Sol 6.2 required more directional guidance; its initial UI designs showed no improvement over 6.1 and repeatedly provided incorrect configuration files, forcing the team to verify project startup capabilities one by one.

CodeRabbit's conclusion is to hold off on switching. Astra's additional catches came with a substantial increase in comments, and the integration still needs comparison against fixed routing configurations; Sol's marginal gains came with more comments and higher costs. The authors suggest reading incremental findings before switching and retesting the issues each configuration stopped catching. While OpenAI claims new models can complete demanding tasks with fewer output tokens, actual review costs depend on delegated calls and retry counts; lower token unit prices alone do not establish lower review costs.

Sources:CodeRabbit

Research

JEDEC Previews DDR6 Standard: Module Length Shrinks to 74mm, Bandwidth Increases Over 50%

Material
Verified 2026-11-20 10:40 GMT+8

The semiconductor standards body JEDEC released a preview of its next-generation DRAM standard, "DDR6," on November 18, 2026. The standard is currently under development, and the features shown reflect current plans that may change during the approval process.

In terms of performance, DDR6 RDIMMs will start at data rates of 12.8 GT/s, aiming for speeds exceeding 14.4 GT/s. The data bus width expands from 64 DQ in DDR5 to 96 DQ, a 50% increase. This architectural adjustment aims to deliver more than a 50% bandwidth increase at introduction through wider channels and higher frequencies.

For hardware designers, the most significant change lies in physical dimensions. The DDR6 module length is set at 74 mm, breaking the 133.35 mm (5.25-inch) standard that has persisted for about 30 years since the SDRAM era. This miniaturization not only improves memory density within systems but may also blur the lines between desktop and laptop modules, as 74 mm is only slightly larger than the current DDR5 SO-DIMM's 69.6 mm.

DDR6 adopts a pseudo-channel architecture to divide memory access and doubles the bank count to optimize refresh operations and data bus efficiency. Additionally, individual DRAM die densities will support up to 128 Gb, with further capacity gains achieved through optimized stacking within packages.

Regarding reliability and energy efficiency, DDR6 supports multiple system-level error-correcting code (ECC) implementations, allowing designers to balance cost, power, and reliability. The interface is optimized for high-capacity modules, explicitly targeting general-purpose computing as well as AI inference and agentic AI applications that demand greater capacity and performance close to the CPU.

Sources:jedec.org

Research

US House Panel Seeks Microsoft Briefing on Reported ByteDance Azure OpenAI Deal

Material
2026-11-20 09:58 GMT+8

The US House Select Committee on the Chinese Communist Party requested a briefing from Microsoft on November 18, following reports by technology media outlet The Information that the company is close to a $500 million deal allowing ByteDance to use OpenAI models through the Azure cloud platform.

Committee Chairman John Moolenaar, a Republican representative from Michigan, stated that Microsoft should not sell AI tools to ByteDance. He noted that US companies helping China catch up in the AI race is a serious concern, and that selling AI services to ByteDance is unacceptable given Microsoft's status as a Pentagon contractor.

According to reports, OpenAI itself restricts most Chinese enterprises from directly using its models, but Microsoft holds the rights to sell OpenAI technology and reportedly does not impose the same restrictions, giving ByteDance access to models it could not otherwise obtain via Azure. These details remain part of ongoing negotiations, and neither company has announced a completed transaction.

A Microsoft spokesperson responded that the company takes the misuse of AI services seriously and commits to enforcing terms of service and complying with applicable laws, but declined to comment on confidential customer negotiations. This dispute highlights a shift in Washington's regulatory focus from chip exports to cloud-based model services.

Sources:INSIDE

Research

European Commission Confirms Cyberattack Using AI Tool ARTEX; Some Internal Systems Still Down

Material
2026-11-20 09:53 GMT+8

The European Commission confirmed on Thursday (November 19) that it detected a cyberattack using AI tools on October 22. This marks the first time the EU's executive branch has publicly acknowledged being targeted by such automated offensive frameworks. As of now, some internal systems remain unavailable, and the extent of potential employee data leakage is still under investigation.

According to Politico and subsequent Commission statements, the attack affected several critical internal systems, including Sysper (the HR platform handling salaries, leave, and timesheets), MyPMO (medical expense reimbursement), and internal translation services. The attackers utilized ARTEX, an open-source AI penetration testing framework released by developer "Autumn-27" in July 2026. ARTEX uses large language models (LLMs) to autonomously execute vulnerability scanning, reconnaissance, exploitation, and privilege escalation.

The Federal Bureau of Investigation (FBI) issued a security alert on November 13 stating that ARTEX had been weaponized by threat actors since August, targeting sectors including finance, government, and technology. The FBI noted that some intrusions progressed from initial targeting to privileged access in under three hours. Although the original developer removed the GitHub repository on October 8, the code continues to proliferate. Investigations by Cloudflare and Spectrum Security indicate that ARTEX deployments and modified derivatives are increasing across the open web.

The EU cybersecurity agency CERT-EU had warned as early as November 5 that ARTEX might be used against EU institutions but did not name specific victims. The Commission's confirmation closes this gap. Spokesperson Thomas Regnier stated that while no evidence of mass data exfiltration has been found yet, investigators are verifying whether sensitive information such as names, salary grades, home addresses, and contact details were accessed. By November 19, Sysper had been restored, but the medical system JSIS remained offline.

Sources:iThome

Research

Omid Ashtari, Founder of UK AI Gaming Companion Kult: Routing Decisions by Confidence Level with Lightweight Models Answering First and Escalating When Uncertain

Material
2026-11-20

An AI gaming companion sitting next to a player must simultaneously satisfy three contradictory requirements: answers must be fast (games are real-time), accurate (each game has its own items, systems, and vocabulary), and low-cost (an always-on companion handles multiple turns per session, each of which must be cheap enough for a consumer product). Omid Ashtari, founder and CEO of London-based startup Kult, explained in a Google Cloud blog how they solved this trilemma using a model routing architecture.

The core idea is not to select a single model, but to let different tiers of models do their respective jobs. The lightest tier, Gemini Flash, handles high-frequency repetitive work—Kult initiates thousands of small calls for every game added. Gemini Flash runs every conversational turn: it determines what the player is asking and writes an answer based on Kult's self-built game knowledge base, fast enough to keep up with the pace of play. Meanwhile, the newest and heaviest Gemini models handle rare, deep reasoning tasks performed once per game. Ashtari noted that the latest Flash models have achieved a balance of price, speed, and quality that they could not find elsewhere, which is precisely why an always-on gaming companion can function as a viable consumer product.

However, tiering alone is insufficient; a mechanism is needed to decide when to upgrade. Kult uses a decision-specific model named Jev to handle such judgments: it does not write content, but returns a choice along with a calibrated confidence score. Cases with high confidence are resolved in a fraction of a second at minimal cost, while uncertain cases are escalated to larger Gemini models. Ashtari explained that much of the work in the system is inherently about decision-making rather than writing—it needs to understand which game the query concerns, whether the question requires a lookup, and whether the answer covers the key points. These decisions do not require large models, but they do need to be reliable.

For this routing to work, there is a prerequisite: answer accuracy does not come from the model's memory, but from Kult's own structured knowledge base. Games may change content with every patch update, and the training cutoff dates of general-purpose models mean they can be months out of date. Kult automatically builds structured knowledge bases for every covered game and keeps them updated; the model's job is to reason over this knowledge rather than rely on memory from training data. Ashtari's design separates factual sources from reasoning capabilities: the knowledge base provides facts, while the model handles reasoning and expression.

Kult currently covers more than 100 games, and adding new games is an automated process rather than a manual project. This allows their small team to target any game players pick up, including new releases—where most gameplay occurs in the first few weeks after launch. The company is currently in the alpha stage, gradually opening access to more players.

It is important to emphasize that this is Kult's self-reported architectural description. The Jev model is not public, so the quality of its confidence calibration cannot be independently verified; the product is in the alpha stage, and there is no public performance data proving that this routing strategy remains effective at scale. However, the design logic provided by Ashtari—using the smallest model capable of doing the job correctly, and using a decision model to determine when to escalate—offers reference value for any team building multi-turn conversational consumer AI products.

Sources:Google Cloud Blog

Research
Next →
Sign in / Sign up
ReadingResearchRadarInvestment framework