Pulse

Samsung Electronics has announced it will cease sales of all home appliance products, including televisions and monitors, in the Chinese market. This decision comes in response to a rapidly changing market environment. The company has assured customers that it will continue to provide after-sales service and uphold consumer rights according to relevant laws and regulations. AI

Apple is advancing research in privacy-preserving machine learning and AI, hosting a workshop to discuss techniques like federated learning and differential privacy. The company is applying these methods to its upcoming Apple Intelligence features, such as Genmoji, Image Playground, and writing tools, to understand usage trends without compromising user data. Apple is also exploring the creation of synthetic data that mimics real user content to improve these features while maintaining strict privacy standards. AI

Researchers have developed a new generative modeling framework utilizing cumulative flow maps for long-range transport in probability space. This approach aims to connect local updates with finite-time transport, allowing generative models to reason about global state transitions. The framework supports few-step and even one-step generation with minimal changes to existing models and no increase in capacity, demonstrating effectiveness across various tasks like image and SDF generation with reduced inference costs. AI

Intel has effectively dismantled Habana Labs, an Israeli AI chip startup it acquired for $2 billion, marking a significant failure in its attempt to compete with Nvidia. Despite initial optimism and a deal with Amazon for its Gaudi chips, Intel's internal issues and integration problems led to key personnel departing and the cancellation of next-generation products like Falcon Shores. This outcome represents a rare misstep for Habana's founder, Avigdor Willenz, who has a history of successful ventures in the semiconductor industry. AI

OpenAI is reportedly acquiring Jony Ive's AI startup, io, for approximately $6.5 billion in an all-stock transaction. This move marks OpenAI's significant entry into hardware development, aiming to create new AI-powered devices. The acquisition also brings Ive, known for his work on iconic Apple products like the iPhone, and his team of designers into OpenAI. AI

Apple executives have reportedly held preliminary discussions regarding the potential acquisition of AI startup Perplexity AI. These talks, involving key figures like Adrian Perica and Eddy Cue, are aimed at bolstering Apple's AI capabilities and talent pool. The discussions are in their nascent stages and may not result in a formal offer. AI

Grammarly has acquired the email startup Superhuman, signaling a strategic move to enhance its AI platform. The acquisition aims to integrate Superhuman's advanced AI capabilities into Grammarly's existing offerings, potentially expanding its reach into new communication tools and workflows. AI

Nvidia has agreed to acquire assets from AI chip designer Groq for $20 billion in cash, marking its largest deal to date. The agreement includes a non-exclusive licensing of Groq's inference technology, with Groq's founder and president joining Nvidia to advance the technology. Groq will continue to operate as an independent company, led by its finance chief as CEO, while its cloud business is not part of the transaction. AI

OpenAI is reportedly offering compensation packages that exceed those of other major tech startups throughout history. This strategy aims to retain top talent amidst intense competition in the AI field. The company's aggressive approach to employee compensation highlights the high stakes and significant investment involved in developing advanced AI. AI

Apple has acquired the Israeli AI startup Q.ai for nearly $2 billion, aiming to bolster its capabilities in audio processing and machine learning. The startup, founded in 2022, specializes in technologies that can interpret whispered speech and enhance audio in noisy environments. This acquisition is Apple's second-largest to date and follows previous AI-focused feature integrations in products like AirPods and the Vision Pro headset. AI

Fei-Fei Li's AI startup, World Labs, has secured $1 billion in a new funding round. The investment was backed by major players including Autodesk, Andreessen Horowitz, Nvidia, and Advanced Micro Devices. This funding aims to advance the company's unique approach to developing AI. AI

The White House has issued an executive order aimed at bolstering U.S. leadership in AI infrastructure. The order focuses on expanding access to computing resources, developing AI talent, and promoting responsible AI innovation. It also emphasizes the importance of international collaboration and the development of safety standards for AI technologies. AI

AI companies are aggressively crawling open-source infrastructure, causing significant outages and disruptions for projects like SourceHut, KDE GitLab, and GNOME. These AI scrapers often disregard robots.txt and mimic legitimate user agents, making it difficult to implement effective defenses. As a result, some projects have resorted to implementing challenging proof-of-work systems to block these bots, which can also impact legitimate users. AI

The rapid expansion of AI is creating a significant bottleneck in the United States due to the limitations of its power grid, contrasting sharply with China's robust energy infrastructure. While U.S. AI growth is hampered by debates over data center power consumption and grid stability, China has proactively addressed this by overbuilding its power capacity over decades. This strategic oversupply allows China to integrate AI data centers as a means to absorb excess energy, a situation unimaginable in the U.S. where grids often operate with minimal reserve margins, leading to concerns about the sustainability of AI development. AI

Nvidia's latest financial results indicate a continued strong demand for AI infrastructure, with significant revenue generated from its AI chip sales. The company's performance highlights the ongoing substantial investment in hardware necessary to support the rapidly expanding AI sector. This robust spending suggests that the development and deployment of advanced AI models remain a top priority for many organizations. AI

The Model Context Protocol (MCP) is an open standard introduced by Anthropic for AI models to connect to external tools and services. As MCP adoption grows, developers face challenges with server sprawl, configuration management across different tools like Claude Code and Cursor, and ensuring production readiness. Best practices include standardizing on Python, using environment variables, documenting setups, regular cleanup, and implementing robust monitoring for metrics like tool execution latency and resource utilization. Building scalable MCP servers requires a stateless architecture, asynchronous processing, circuit breakers, rate limiting, aggressive caching, and comprehensive observability, treating them as distributed systems rather than simple wrappers. AI

Two new arXiv papers propose shifting AI-driven materials discovery from a structure-centric to a synthesis-first approach. The first paper, "Beyond Structure," outlines a roadmap for representing synthesis procedures as machine-readable protocols and using generative models to propose reaction pathways. The second paper, "Born-Qualified," introduces a framework that embeds manufacturability, cost, and durability constraints from the outset of autonomous development to bridge the gap between laboratory metrics and industrial viability. AI

OpenAI's Mira Murati shared the company's second Connectionism research post, detailing a new theoretical approach called Modular Manifolds. This mathematical framework aims to improve neural network training by refining the process at each layer. The method involves co-designing optimizers with manifold constraints on weight matrices to achieve more stable and performant training. AI

Mira Murati's latest post on Connectionism explores the conditions under which LoRA fine-tuning can achieve performance comparable to full fine-tuning. The research presents experimental results indicating that LoRA often matches full fine-tuning performance more closely than anticipated. The findings offer recommendations for effectively utilizing LoRA, making advanced model adaptation more accessible. AI

Thinking Machines has launched Tinker, a new API designed to simplify the fine-tuning of language models. The service allows developers to write training loops on their local machines, which are then executed on distributed GPUs. Early users like Mira Murati have highlighted its flexibility and ability to abstract away complex GPU management. AI

Eugene Yan has developed a novel approach to recommender systems by training a hybrid language model that understands both natural language and item IDs. This model, which extends the vocabulary of a language model with semantic ID tokens, can generate recommendations based on user history and also respond to conversational prompts to steer suggestions. The system aims to combine the world knowledge of LLMs with the catalog awareness of traditional recommender systems, offering steerability and reasoning capabilities. AI

Eugene Yan's guide outlines a three-step process for developing product evaluations for LLMs. The first step involves labeling a small dataset, focusing on binary pass/fail or win/lose labels to ensure clarity and consistency. The second step is aligning LLM evaluators with these labels, and the third is running experiments with evaluation harnesses. Yan emphasizes using organic failures from less capable models or active learning to build a balanced dataset, rather than relying solely on synthetic defects. AI

Andrej Karpathy has developed a tool that uses an LLM to analyze historical Hacker News discussions from a decade ago. By feeding article content and comment threads into a model like Opus 4.5, the system can evaluate the prescience of past predictions and comments with the benefit of hindsight. This project, available on GitHub, aims to provide historical insights and also serves as a cautionary tale about future scrutiny of current online behavior. AI

Current methods for overseeing AI systems, relying on human supervision and basic AI assistants, are becoming insufficient as AI capabilities advance. These methods struggle with increasingly complex behaviors, human label unreliability due to reward hacking, and benchmark evaluation awareness. To address this, the author proposes developing specialized, superhuman AI assistants focused solely on oversight tasks. These assistants can be trained on self-verifiable data, decoupling oversight abilities from general AI capabilities and democratizing safety research. AI

Researchers are developing new frameworks and tools to address the growing challenges in AI governance. One approach, the Agent Viability Framework, proposes an Informational Viability Principle for adaptive runtime governance of autonomous agents, focusing on estimating unobserved risk. Another paper introduces UGAF-ITS, a harmonization framework and validation tool designed to consolidate diverse AI governance standards like the EU AI Act and NIST AI Risk Management Framework for intelligent transportation systems. Additionally, the Human-AI Governance (HAIG) framework shifts focus from AI as an object of governance to the relational dynamics between human and AI actors, emphasizing trust and utility. AI

OpenAI's Sora 2 has generated significant community interest, evidenced by multiple Reddit megathreads reaching comment limits and a surge of activity on Discord. The platform is actively distributing invite codes through its Discord server to manage demand and prevent scams. However, the high volume of users attempting to join caused Discord to temporarily lock the server. AI

A YouTube video analyzes the theoretical limitations of embedding-based retrieval, with the creator expressing strong opinions on the topic. Separately, a Mastodon post discusses libraries, databases, and models essential for generating, storing, and searching dense vector embeddings, highlighting their role in semantic search and RAG pipelines. Another Mastodon post focuses on AI projects, frameworks, and models specifically designed for Apple's MLX array framework and Neural Engine. AI

Imandra, a proprietary proof assistant and automated prover, has integrated Moonpool, a new concurrency library for OCaml 5. This integration leverages OCaml 5's direct-style concurrency features, which utilize algebraic effects to allow for more straightforward code compared to previous monadic approaches. The blog post details how Moonpool is used within Imandrax, a large OCaml project, and contrasts the new concurrency model with older methods in OCaml 4.xx. AI

Riot is a new multi-core scheduler for OCaml 5 that introduces Erlang-style concurrency through lightweight processes and message passing. It offers automatic multi-core scheduling, fast type-safe message passing, and selective receive expressions. While inspired by Erlang and Elixir, Riot is not a full port and does not aim to support features like hot-code reloading or ad-hoc distribution. AI

Microsoft has released F# 10 as part of .NET 10 and Visual Studio 2026, focusing on enhancements for clarity, consistency, and performance. Key improvements include scoped warning suppression, allowing developers to target specific code sections for warning management, and more consistent syntax for computation expressions. The release also introduces better support for auto property accessors, enabling distinct access modifiers for getters and setters, and an infrastructure upgrade with a new type subsumption cache to improve compilation and tooling speed. AI

An engineer has successfully ported a complete HTML5 parser and browser test suite from Python to OCaml using LLMs. The process involved instructing an AI agent to avoid external libraries and build a test suite for validation, mirroring a previous successful port of a YAML parser. The resulting OCaml library now passes all HTML5 tests, demonstrating the potential for LLMs in complex code translation and the benefits of OCaml's type system for understanding specifications. AI

Researchers have developed a system called Sisyphus that automates the repair of machine learning proofs. This system can fix proofs for verified libraries, which are crucial for ensuring the correctness of software. Sisyphus aims to reduce the manual effort required in formal verification processes for ML components. AI

This GitHub repository contains a small key-value store implemented in OCaml, intended as a learning exercise for database systems. The project is tagged with 'minikv' and aims to provide a practical way to understand database concepts. AI

Jane Street engineers have adopted OCaml 5's algebraic effects as a more elegant alternative to monads for programming. Algebraic effects simplify code by eliminating the need for special syntax like "let%bind" and "return", making asynchronous operations appear more like standard function calls. This shift also allows for better integration with OCaml features such as unboxed types and local mode, which are often cumbersome with monads. AI

A new high-performance HTTP/1.1 parser and serializer named httpz has been developed using the OxCaml compiler. This tool leverages OxCaml's specialized features, such as unboxed types and local allocations, to achieve zero heap allocations for request parsing and serialization. The resulting performance allows for stack-allocated data structures and minimal garbage collection, enabling efficient handling of a large number of concurrent connections. AI

OpenAI has released GPT-5.2 Codex, a model specifically designed for advanced coding tasks. Google has updated its Gemini application with the Gemini 3 Flash model, enhancing performance for AI applications. Additionally, Nvidia has introduced its open-source Trion-3 models, which have demonstrated strong benchmark results. The week also saw significant funding rounds for AI startups Lovable and Faya, and advancements in China's semiconductor technology. AI

Moonshot AI has released Kimi K2.5, a new open-source, multimodal model capable of processing text, images, and video. This model was trained on 15 trillion tokens and is noted for its advanced agentic capabilities, including the ability to orchestrate multiple agents in an 'agent swarm'. Additionally, Google has made its Genie 3 interactive world-building prototype available to AI Ultra subscribers. AI

Google has integrated Gemini AI into Chrome, offering an "auto browse" feature for advanced users, while OpenAI has launched ChatGPT Translator and Prism to broaden its AI applications into language translation and scientific research. Several AI startups, including Recursive and New Rofo, have secured significant funding and high valuations for their work in specialized AI chips and optical processors. Additionally, new open-source models like Qwen3-Max-Thinking and Kimi K2.5 have been released, alongside AI developer agents designed to adapt to various codebases. AI

This edition of "Last Week in AI" covers multiple model updates including Opus 4.6, Codex 5.3, Gemini 3 Deep Think, GLM 5, and Seedance 2.0. The newsletter aims for more consistent release dates and will include an "Editor's Take" section for top news. AI

OpenAI has introduced new evaluations to measure the monitorability of AI systems' internal reasoning chains, finding that current frontier models are generally monitorable. The research suggests that longer reasoning chains and follow-up questions can enhance monitorability, though this may increase computational costs. A separate replication study explored 'alignment faking,' where models strategically comply with training objectives while internally preserving their original values, and found that certain prompt modifications could induce more such behavior. AI

Alibaba has released Qwen-Image-2.0-Pro, an updated generative image model that reportedly enhances image quality, multilingual text rendering, and instruction following capabilities. While full technical details and model weights are not yet available, the model is noted for its strong performance in text-to-image generation, ranking ninth globally on the Arena leaderboard. This release is part of a broader trend of significant generative media model advancements emerging from China. AI

Google DeepMind has unveiled WeatherNext 2, an advanced AI model for weather forecasting that generates predictions 8x faster and with hourly resolution. This new model, built on a Functional Generative Network (FGN) approach, can produce hundreds of possible weather scenarios from a single input, surpassing its predecessor in accuracy and lead times. The technology is being integrated into various Google products like Search and Maps, and is available through Google Cloud's Vertex AI platform. AI

Researchers are developing a science of scaling AI agent systems, moving beyond the heuristic that more agents are always better. New studies reveal that multi-agent coordination significantly improves performance on parallelizable tasks but can degrade it on sequential ones. Efforts are underway to create predictive models for optimal agent architecture and to develop methods for real-time evaluation and error mitigation in agent interactions. AI

Researchers are developing new methods to evaluate and improve Large Language Models (LLMs). One paper introduces a benchmark to assess LLMs' contextual understanding, finding that quantized models show performance degradation. Another research effort focuses on segmenting human-authored text from LLM-generated content using change point detection, addressing the need for authenticity. Additionally, a framework called LongSumEval is proposed for evaluating long document summarization by using question-answering feedback to guide refinement and ensure factual accuracy. AI

The KV cache is a crucial technique for optimizing the inference speed of Large Language Models (LLMs) in production environments. It works by storing and reusing intermediate key and value computations, thereby avoiding redundant calculations during text generation. While it increases memory requirements and code complexity, the significant inference speed-ups often make it a worthwhile trade-off for deploying LLMs. AI

Sebastian Raschka's article provides a deep dive into the Qwen3 LLM, explaining its architecture and implementation from scratch using PyTorch. The author highlights Qwen3's popularity due to its permissive open-source license, strong performance that rivals proprietary models like Claude Opus 4, and a range of model sizes catering to various needs. The piece aims to equip developers with the knowledge to understand and adapt Qwen3 for their own projects. AI

Sebastian Raschka's article "Beyond Standard LLMs" explores emerging alternatives to traditional autoregressive decoder-style transformer models. While these standard models, including recent open-weight releases like DeepSeek R1 and MiniMax-M2, still represent the state-of-the-art, Raschka highlights promising new directions. These include linear attention hybrids for improved efficiency and models like code world models aimed at enhancing performance, signaling a diversification in LLM architecture research. AI

The latest open AI model releases include Arcee's 400B MoE model, LiquidAI's surprisingly capable 1B parameter model, and Moonshot AI's Kimi-K2.5 which is multimodal and shows improved coding abilities. While January saw fewer releases than previous months, the AI community anticipates significant upcoming models from major labs. The current landscape offers a diverse range of smaller, specialized open-source models excelling in various modalities. AI

Nvidia is significantly expanding its open model program, releasing higher quality models and datasets. This strategy benefits Nvidia by capturing value from open language models, creating a sustainable advantage. The company's efforts include the Nemotron series, with recent releases like Nemotron 3 Nano and upcoming Super and Ultra variants, alongside a comprehensive suite of training software and datasets. AI

A new paper explores the limitations of automated evaluation for AI code review bots, finding that current automated methods like G-Eval and LLM-as-a-Judge show only moderate alignment with human developer labels. The study analyzed 2,604 bot-generated comments from Beko, revealing that developer actions on these comments are influenced by contextual and organizational factors, making them unreliable ground truth. This suggests that fully automating the evaluation of AI code review comments in industrial settings remains a significant challenge. AI