MCP-B: A Protocol for AI Browser Automation

MCP-B is a protocol that enables AI assistants to directly interact with a website's functions, allowing for instant and secure automation with zero configuration, by adding just 50 lines of code. This approach replaces traditional browser automation methods, which can be slow and brittle, with a more reliable and efficient solution that uses existing browser authentication and respects existing permissions.

Read more

Springer Nature book on machine learning is full of made-up citations

A Springer Nature book on machine learning, "Mastering Machine Learning: From Basics to Advanced", has been found to contain numerous made-up citations, with two-thirds of the 18 checked citations either not existing or having substantial errors. The book's author, Govindakumar Madhavan, has not confirmed whether he used a large language model like ChatGPT to generate text for the book, but the publisher, Springer Nature, is investigating the matter and emphasizes the importance of human oversight and declaration of AI use in submissions.

Read more

Hugging Face just launched a $299 robot that could disrupt the robotics industry

Hugging Face, a $4.5 billion AI platform, has launched Reachy Mini, a $299 desktop robot designed to bring AI-powered robotics to millions of developers worldwide, marking a significant move to democratize robotics development and challenge the traditional closed-source, high-cost model. The robot, which can be programmed in Python and integrated with Hugging Face's platform, aims to make robotics more accessible and open-source, allowing developers to create and share new applications and potentially creating a vast ecosystem of robotics apps.

Read more

Why I don't ride the AI Hype Train

The author is skeptical of the AI hype surrounding ChatGPT and other large language models, citing concerns over how these models are trained using copyrighted content without permission and the significant environmental impact of the data centers used to train them. The author also argues that the actual use of these models is often disappointing and can have negative consequences, such as students using them to cheat on homework or lawyers relying on them for legal advice, highlighting the need for a more nuanced understanding of the limitations and risks of AI technology.

Read more

That white guy who can't get a job at Tim Hortons? He's AI

TikTok has removed a series of AI-generated videos featuring a white man named "Josh" who complains about the difficulty of getting a job in Canada, making racially charged statements about immigrants. The videos, created by AI firm Nexa, were part of a marketing campaign and have been criticized as deceptive and unethical, with TikTok taking them down for violating its community guidelines due to lack of clear labeling as AI-generated content.

Read more

The Cost of an Image: The Energy Consumption of AI Image Generation

The environmental impact of AI image generation is a growing concern, with a recent study finding that 17 state-of-the-art image generation models vary drastically in energy consumption, with up to a 46x difference. The study's results show that factors such as model architecture and image resolution affect energy consumption, but surprisingly, improving image quality does not always increase energy consumption, and some models can produce high-quality images while being energy efficient.

ArXiv

Do AI Tutors Empower or Enslave Learners?

The increasing use of AI in education can have negative effects, such as cognitive atrophy and loss of agency, if not used intentionally and transparently. To avoid these risks, educators should consider the students' perspectives and implement strategies that ensure AI supports learning without undermining core educational goals, ultimately empowering rather than diminishing the learner.

ArXiv

Amazon gets serious with AI Safety

Nova Premier, Amazon's most advanced multimodal foundation model, has undergone a comprehensive evaluation of its critical risk profile, targeting high-risk domains such as CBRN and cyber operations. The evaluation found that Nova Premier is safe for public release, meeting the commitments made at the 2025 Paris AI Safety Summit, and Amazon will continue to enhance its safety evaluation and mitigation pipelines as new risks emerge.

ArXiv

MemOS: A Memory OS for AI System

Large Language Models (LLMs) are hindered by their lack of well-defined memory management systems, limiting their ability to track user preferences and update knowledge over time. The proposed MemOS, a memory operating system, addresses this challenge by unifying the representation, scheduling, and evolution of different memory types, enabling cost-efficient storage and retrieval, and laying the foundation for continual learning and personalized modeling.

ArXiv

Fun with flags: How Compilers Break and Fix Constant-Time Code

Developers' efforts to prevent timing side-channel attacks through constant-time programming can be undermined by compiler optimizations that reintroduce leaks. This paper identifies the specific optimization passes in compilers GCC and LLVM that cause these leaks and proposes a practical solution: disabling selected passes via compiler flags, which significantly reduces leakage with minimal performance overhead.

ArXiv

Biomni: A General-Purpose Biomedical AI Agent

Biomni is a general-purpose biomedical AI agent that integrates large language model reasoning with retrieval-augmented planning and code-based execution to autonomously execute research tasks and generate testable hypotheses. The open-science initiative provides a software environment, web interface, and invites community contributions to enhance research productivity and accelerate science, with contributors being acknowledged and potentially invited as co-authors on upcoming papers.

Repo

Reuse non-prefix KV Cache and speed up RAG by 3X with LMCache

LMCache is a caching system, but details about its functionality and purpose are not provided in the given text. The provided text only mentions that it is the example page for LMCache, without offering any further information about its features or usage.

Repo

Show HN: PromptDrifter – Catch LLM prompt drift before it breaks prod

PromptDrifter is an open-source platform that helps catch "prompt drift" in Large Language Models (LLMs) by validating their responses against expected outcomes, ensuring consistent and reliable applications. It provides a simple, command-line driven tool to integrate LLM response validation into development and CI/CD workflows, supporting various LLM providers and models through its adapter system.

Repo

Show HN: A protocol for LLM client to render UI components

The Language Model User Interface (LMUI) Protocol is a standardized method for enabling rich, interactive user interfaces within conversational AI experiences, addressing limitations of plain text interactions such as ambiguity, inefficiency, and limited experience. The LMUI protocol allows language models to request information using UI components directly within the chat interface, enabling richer, more efficient, and more intuitive conversational flows through a simple, extensible contract.

Repo

Show HN: AI-Friendly Toolchain – Dev Tools for Working with LLMs

This is a curated list of AI-adjacent developer tools, including scripts, visualizers, templates, and utilities that make working with Large Language Models (LLMs) and AI systems smoother. The list covers various categories such as prompt engineering, context tools, cost tracking, and utilities, and is open to contributions from developers who have found other useful AI-adjacent tools.

Repo