Pulse

Researchers have developed a novel approach to accelerate machine learning on Field-Programmable Gate Arrays (FPGAs) using Kolmogorov-Arnold Networks (KANs). This method aims to achieve ultrafast inference and online learning by implementing neural networks directly as digital logic, bypassing the overhead associated with traditional processors like GPUs. The work, detailed in two papers, focuses on efficient evaluation and spline locality for KANs on FPGAs, addressing the need for ultra-low latency and high hardware efficiency in specialized applications. AI

Cognition AI has launched FrontierCode, a new benchmark designed to evaluate the quality of AI-generated code beyond mere correctness. This benchmark was developed with input from over 20 open-source developers and focuses on whether code would be accepted into real-world production codebases. Early results show that even top-tier models like Anthropic's Claude Opus 4.8 struggle, achieving only a 13.4% score on the most challenging subset, indicating a significant gap in producing high-quality, maintainable code. AI

A new study published in PNAS suggests transitioning corn-for-ethanol farmland to solar energy production could significantly boost the US's energy output while reducing ecological pressures. Researchers found that converting just 3.2% of land currently used for corn ethanol could generate the same amount of energy as all current corn ethanol farming. This shift could also decrease fertilizer use and irrigation needs, while potentially offering farmers higher earnings than crop cultivation. AI

Anthropic is reportedly developing two new models, codenamed Mythos and Capybara. Details about these models are scarce, but their existence suggests ongoing advancements in Anthropic's AI capabilities. The information emerged from a leaked internal document or presentation. AI

The Free Software Foundation (FSF) has commented on the settlement in the Bartz v. Anthropic copyright infringement lawsuit. This class action suit alleges Anthropic used copyrighted materials from datasets like Library Genesis to train its large language models. While a court initially suggested training LLMs on these works might be fair use, the FSF, holding copyrights to works like "Free as in Freedom," is seeking user freedom as compensation, advocating for transparency in LLM training data and code. AI

Mog is a new programming language designed for AI agents to modify themselves safely and efficiently. It is statically typed and compiled, allowing AI agents to write, compile, and load Mog programs as plugins with controlled function access. The language emphasizes security through its Rust-based compiler and explicit type conversions, aiming to enable agents to extend their own capabilities. AI

Now I Get It is a new tool that transforms scientific papers into interactive webpages. Users can upload a PDF and receive an explanation tailored to different audiences, including technical, general, and kid-friendly versions. The service offers free credits for initial users and has a file size limit for uploads. AI

A new benchmark called the "Car Wash Test" reveals that many leading AI models struggle with basic reasoning. When asked whether to walk or drive 50 meters to a car wash, 42 out of 53 tested models incorrectly suggested walking. Even top-tier models like Claude Sonnet 4.5 and GPT-5.2 failed the test on a single run. Consistency tests showed further degradation, with only five models reliably answering correctly across ten attempts, highlighting a significant gap in practical reasoning capabilities. AI

A developer has created a self-contained Unix-like environment within a single 420KB HTML file, accessible in a browser without a server. This environment includes a shell, Git, Node.js, a C compiler, SQLite, Python, and integrates with the Claude Code API for AI-assisted coding. Separately, another developer built an automated pipeline using Node.js and Python to process large datasets of AI interaction logs, identifying and implementing new user-defined skills for AI platforms. AI

A developer has created a tool that uses Anthropic's Claude Code to analyze books and identify thematic connections. The project, called "Useful Lies," visualizes these relationships, offering insights into concepts like self-deception, innovation, and the dynamics of mega-projects. The tool aims to automatically discover and present thematic links across a collection of texts, making complex ideas more accessible. AI

A new open-source CLI tool called Continuous Claude has been developed to automate complex coding tasks by running Anthropic's Claude Code model in a persistent, iterative loop. This tool addresses the limitation of current AI coding assistants that often stop after a single task, enabling multi-step projects to be completed autonomously. By maintaining context across iterations and integrating with GitHub's CI/CD workflows, Continuous Claude can autonomously create branches, generate commits, push changes, monitor checks, and merge pull requests, learning and adapting from previous attempts. AI

The Factorio Learning Environment (FLE) has released version 0.3.0, introducing significant advancements for testing AI agents in complex, long-term planning scenarios. This update integrates Claude Code into Factorio, allowing agents to interact programmatically with the game environment without needing the client. New features include a headless renderer for multimodal research and standardization to the OpenAI gym interface, simplifying integration and enabling scalable experimentation. AI

A new class of foundation models called Time-Series Language Models (TSLMs) has been introduced, designed to natively process and reason about temporal data. These models, developed by a team with affiliations to ETH, Stanford, Harvard, and other institutions, aim to bridge the gap between real-world time-series signals and AI-driven decision-making. The project includes both open-source base models and advanced proprietary versions for enterprise applications, envisioning a future where TSLMs enhance fields like healthcare, robotics, and infrastructure. AI

Flywheel AI, a Y Combinator S25 startup, has launched a system for remote teleoperation and autonomy in excavators. Their retrofit solution mechanically actuates existing excavator controls, addressing the lack of electronic interfaces in most hydraulic machines. This enables increased site safety and productivity, while also generating crucial egocentric observation and action data for training autonomous systems. Flywheel is open-sourcing 100 hours of this collected excavator dataset to facilitate research in robot learning. AI

A new guide compiles essential machine learning equations, focusing on their practical application and mathematical foundations. It covers key concepts from information theory, linear algebra, and optimization, including detailed explanations and Python implementations for entropy, cross-entropy, and KL divergence. The resource aims to serve as a handy reference for practitioners, drawing from frequently used formulas and including sections on neural network fundamentals and loss functions. AI

Dyna is a new programming language designed for machine learning researchers, aiming to bridge the gap between mathematical concepts and executable code. It builds upon logic programming paradigms like Datalog and Prolog, introducing features such as flexible execution orders and weighted rules. This allows for concise expression of complex algorithms, including matrix multiplication, the Fibonacci sequence, and neural networks, with minimal code. AI

A programmer explores a childhood mystery surrounding the source code for a BASIC game called Hexapawn. This game, a simplified version of chess, was featured in an old programming book. The author delves into the game's DATA statements, which initially appeared as incomprehensible sequences of numbers, and seeks clarification from Claude.ai to understand their function within the game's logic. AI

Glowstick is a new Rust crate designed to enhance tensor manipulation by integrating shape checking directly into the type system. This approach aims to make tensor operations safer and more intuitive, particularly for developers working with machine learning frameworks. The project, currently in its pre-1.0 phase, offers features like dynamic dimension support and improved error messages, with plans to align with ONNX operations. AI

Researchers have introduced a new framework called "The Illusion of Thinking" to better understand the reasoning capabilities and limitations of Large Reasoning Models (LRMs). This framework utilizes controllable puzzle environments to analyze the internal reasoning traces of LRMs, moving beyond traditional evaluations that focus solely on final answer accuracy. Experiments revealed that LRMs experience a complete accuracy collapse at high problem complexities and exhibit a peculiar scaling limit where reasoning effort decreases despite sufficient computational resources. AI

Researchers have developed SeedLM, a novel post-training compression technique for large language models that utilizes pseudo-random generator seeds to encode model weights. This method aims to reduce the high runtime costs associated with LLMs by generating weight matrices on-the-fly during inference, thereby decreasing memory access and improving speed for memory-bound tasks. SeedLM achieves this by trading compute for fewer memory accesses and notably does not require calibration data, generalizing well across diverse tasks and maintaining accuracy comparable to FP16 baselines even at significant compression levels. AI

A developer is creating a versatile OCR pipeline designed to extract structured data from complex educational materials for machine learning training. The system, which supports multilingual text, mathematical formulas, tables, and diagrams, aims to achieve over 90-95% accuracy on academic datasets. It generates AI-ready outputs in JSON or Markdown, including semantic annotations for visual content, and is built using various tools like Google Vision API and OpenAI API. The project's public release has been delayed due to the developer's academic commitments but is expected once the system is finalized. AI

A new open-source framework called FormalVerifML has been released, utilizing Lean 4 for the formal verification of machine learning models. This tool aims to provide mathematically rigorous proofs of properties like robustness, fairness, and safety for high-stakes applications. It supports large-scale models, including transformers and vision models, with features for enterprise use and distributed verification. AI

This PDF provides a comprehensive overview of the mathematical foundations essential for computer science and machine learning. It covers topics ranging from linear algebra and calculus to probability and statistics, aiming to equip readers with the necessary quantitative skills for advanced study and research in these fields. The material is structured to build a strong theoretical understanding, enabling practitioners to better grasp and develop complex algorithms and models. AI

Salesforce has released Merlion 2.0, an open-source Python library designed for time series intelligence. The framework offers an end-to-end solution for tasks such as forecasting, anomaly detection, and change point detection. Merlion 2.0 includes a diverse set of models, automated hyperparameter tuning, and practical post-processing rules to enhance model interpretability and reduce false positives. AI

DeepSource has open-sourced Globstar, a static analysis toolkit designed for creating custom code quality and security checkers. The toolkit leverages tree-sitter for parsing code and utilizes AI assistants like ChatGPT and Claude to generate complex queries, simplifying the process for developers. Globstar offers both YAML and Go interfaces, supporting over 20 languages with plans to add C/C++ support. AI

Researchers at Apple have developed ELEGNT, a framework for designing robot movements that blend functional task fulfillment with expressive qualities like intention and emotion. Their work, detailed in a recent paper, involved creating a lamp-like robot and a methodology to generate movement sequences that enhance user engagement, particularly in social contexts. A user study confirmed that expression-driven movements were perceived more positively than purely function-driven ones. AI

A former MIT student reflects on a hardware security research paper he co-authored, "There’s Always a Bigger Fish: A Clarifying Analysis of a Machine-Learning-Assisted Side-Channel Attack." The paper, which demonstrated a machine-learning-assisted side-channel attack executable in web browsers and highlighted how system interrupts can leak user information, has received significant awards. The author discusses the challenges of writing about the research, particularly the dual narrative of ML's potential for attacks and its frequent misapplication, and how the project profoundly influenced his academic and personal path. AI

Researchers are developing AI models to predict and control the behavior of fusion plasma in real-time. These models aim to optimize the process of achieving stable fusion reactions, which is crucial for developing clean energy sources. The project utilizes machine learning techniques to analyze complex plasma dynamics and enable precise manipulation. AI

This paper explores the gap between current AI's language capabilities and its mathematical reasoning abilities. It proposes an information-theoretical approach to developing an AI mathematician, focusing on discovering new conjectures rather than proving existing theorems. The core idea is that a valuable set of theorems should efficiently summarize provable statements and be closely related to many of them. AI

The concept of model homotopy, applying topological ideas to machine learning, suggests that a single model may not fully capture a modeling situation. Instead, a trajectory of fits, parameterized continuously by weights, can offer a richer understanding. This approach can reveal counter-intuitive behaviors, such as linear regression coefficients changing signs multiple times as variables are added, challenging the intuition that coefficients would smoothly interpolate. AI

This article introduces Micrograd.jl, a new automatic differentiation package for the Julia programming language. It aims to fill a gap in comprehensive tutorials for AD in Julia, requiring a solid understanding of both Julia and Calculus. The package is built upon Zygote.jl and ChainRules.jl, offering a different approach to AD compared to Python frameworks like PyTorch by leveraging Julia's functional programming and metaprogramming capabilities. AI

Researchers Bartłomiej Cupiał and Maciej Wołczyk observed a significant performance drop in their neural network trained to play NetHack. The model, which had been consistently scoring around 5,000 points, suddenly began scoring only 3,000 points, a 40% decrease. Despite extensive troubleshooting, including code reversion, software stack restoration, and rebuilding the entire system from scratch, the performance issue persisted. AI

Stein's paradox, a counterintuitive statistical concept, demonstrates that in dimensions three and higher, a better estimate of a Gaussian distribution's mean can be achieved than simply using the drawn sample. The James-Stein estimator, which uses a specific formula involving the sample's magnitude and dimensionality, outperforms the naive approach in terms of mean squared error. This paradox challenges conventional statistical intuition, particularly regarding parameter estimation in higher-dimensional spaces. AI

The Opus 1.5 audio codec has been released with significant machine learning enhancements, marking the first time deep learning is used to process audio signals directly. These new ML-based features, including improved packet loss concealment (PLC) and a novel redundancy transmission method, are designed to be fully compatible with older versions and optimized to run efficiently on standard CPUs. While most users won't notice the performance impact, the ML features are disabled by default and require specific compile-time and run-time flags to activate. AI