News/Open-source
Tesla’s ex-AI chief releases $100 DIY ChatGPT-ish toolkit
AI luminary Andrej Karpathy, former director of AI at Tesla and OpenAI co-founder, has released nanochat—a comprehensive toolkit that lets anyone build their own ChatGPT-like AI assistant for approximately $100. Released on October 13, 2025, this open-source project represents the most accessible path yet for creating a functional large language model (LLM) from scratch. Unlike previous AI-building tools that required extensive technical expertise or substantial financial investment, nanochat provides a complete "full-stack" solution. This means it handles every step of the AI creation process, from initial data processing to deploying a web interface where users can chat with their custom...
read Oct 9, 2025North Carolina’s All Things Open 2025 dedicates 25% of programming to AI
All Things Open returns to Raleigh next week for its 13th annual conference, expecting over 6,000 attendees and featuring more than 200 sessions across 18 tracks at the Raleigh Convention Center. The east coast's largest open source conference is placing unprecedented emphasis on artificial intelligence, with AI-focused sessions comprising nearly 25% of all scheduled talks across three dedicated tracks. What you should know: The conference has evolved into a comprehensive showcase of open source innovation with a heavy AI focus. More than 6,000 attendees are registered, representing 500 companies with 100+ exhibitors participating. AI will dominate programming with over 50...
read Oct 7, 2025Anthropic dishes out open-source Petri tool to test AI models for deception
Anthropic has released Petri, an open-source tool that uses AI agents to test frontier AI models for safety hazards by simulating extended conversations and evaluating misaligned behaviors. The tool's initial testing of 14 leading AI models revealed concerning patterns, including instances where models attempted to "whistleblow" on harmless activities like putting sugar in candy, suggesting they may be influenced more by narrative patterns than genuine harm prevention. What you should know: Petri (Parallel Exploration Tool for Risky Interactions) deploys AI agents to grade models on their likelihood to act against human interests across three key risk categories. The tool evaluates...
readGet SIGNAL/NOISE in your inbox daily
All Signal, No Noise
One concise email to make you smarter on AI daily.
Over Overviews? New browser extension lets users hide Google’s AI search features
A new browser extension called "Bye Bye, Google AI" allows users to completely hide Google's AI Overviews and other AI-powered search features from their results pages. Developed by Avram Piltch, former Editor-in-Chief of Tom's Hardware, the tool addresses growing user frustration with AI-generated summaries that have reached over 2 billion monthly users but face criticism for accuracy issues and cluttering search results. How it works: The extension uses CSS (web styling code) to block AI elements from appearing on Google Search, restoring a cleaner, more traditional search experience.• Users can remove AI Overviews (summaries above web results), the AI Mode...
read Oct 6, 2025Google DeepMind’s CodeMender AI fixes 72 security bugs automatically
Google DeepMind has unveiled CodeMender, an AI agent that automatically fixes software vulnerabilities and proactively rewrites code for better security. Over the past six months, the system has already contributed 72 security fixes to open-source projects, including some with up to 4.5 million lines of code, demonstrating AI's growing capability to address the mounting challenge of software security at scale. How it works: CodeMender leverages Gemini Deep Think models to create an autonomous debugging agent equipped with sophisticated validation tools. The system uses advanced program analysis including static analysis, dynamic analysis, differential testing, fuzzing, and SMT solvers (mathematical problem-solving tools)...
read Oct 2, 2025Linux Foundation launches Newton open-source physics engine for robotics AI
The Linux Foundation has launched Newton, an open-source, GPU-accelerated physics engine co-developed by Disney Research, Google DeepMind, and Nvidia to advance robotics AI development. The platform aims to lower barriers to robotics research by enabling faster, more scalable simulations that help robots learn complex skills and transfer them to real-world applications. What you should know: Newton addresses modern challenges in building generalist robots through advanced simulation capabilities. Built on Nvidia Warp (a programming framework for high-performance computing) and OpenUSD (a universal file format for 3D graphics), the engine delivers GPU-accelerated simulation with a flexible, extensible architecture supporting multiple physics solvers....
read Oct 1, 2025Everyone an Edison: Ex-OpenAI CTO launches Tinker to democratize AI model fine-tuning
Thinking Machines Lab, the heavily funded AI startup cofounded by former OpenAI CTO Mira Murati, has launched Tinker, a tool that automates the creation of custom frontier AI models through fine-tuning. The product represents the company's bet that democratizing access to advanced model customization will be the next major frontier in artificial intelligence development. What you should know: Tinker allows businesses, researchers, and hobbyists to fine-tune cutting-edge AI models without managing complex infrastructure or specialized software tools. Users can currently fine-tune two open source models: Meta's Llama and Alibaba's Qwen through supervised learning or reinforcement learning methods. The tool abstracts...
read Sep 30, 2025DeepSeek cuts AI processing costs 50% with new sparse attention tech
Chinese AI startup DeepSeek has launched DeepSeek-V3.2-Exp, an experimental model that introduces "sparse attention" technology to cut AI processing costs in half while maintaining performance levels. The release builds on DeepSeek's reputation for creating efficient AI systems using fewer resources than traditional approaches, though experts question whether the cost-cutting architecture compromises model reliability and safety. What you should know: DeepSeek's new experimental model represents a significant shift in AI architecture design, focusing on efficiency over raw computational power. The V3.2-Exp model introduces DeepSeek Sparse Attention (DSA), which selectively processes only the most relevant information rather than analyzing all available data....
read Sep 30, 2025If you type it, they will come: Nothing debuts store with apps created by text prompts
Nothing has launched Playground, an app store featuring user-designed, AI-generated applications that can be created through written prompts alone. The London-based smartphone company positions this as the "first step towards an AI-native operating system," though the platform runs on Android and represents more of an interface layer than a true OS replacement. What you should know: Playground is part of Nothing's broader Essential product line, which includes several AI tools already in use. Essential encompasses "all our AI-related products," according to founder and CEO Carl Pei, including an existing AI search tool and Essential Space for organizing voice notes and...
read Sep 29, 2025NVIDIA unveils open-source robotics platform with physics simulation and AI reasoning
NVIDIA unveiled a comprehensive robotics development platform at the Conference on Robot Learning (CoRL) in Seoul, introducing open-source tools designed to accelerate the transition from research labs to real-world robotic applications. The semiconductor giant's latest announcements center on solving fundamental challenges in robotics development: creating realistic simulations for training, enabling human-like reasoning in robots, and bridging the gap between virtual testing and physical deployment. The new platform combines three core technologies that work together as an integrated system. The Newton Physics Engine provides realistic simulation environments where robots can safely learn complex tasks. The Isaac GR00T foundation model serves as...
read Sep 25, 2025Nvidia open-sources Audio2Face AI tool for realistic 3D avatar animations
Nvidia has open-sourced Audio2Face, its AI-powered tool that generates realistic facial animations for 3D avatars based on audio input. The move allows developers worldwide to integrate the technology into games and applications, potentially accelerating the development of more lifelike digital characters across the entertainment and tech industries. What you should know: Audio2Face analyzes acoustic features in voice recordings to automatically generate facial expressions and lip movements for 3D characters. The tool can create animations for both pre-scripted content and real-time livestreams, offering flexibility for different use cases. Developers gain access to the underlying models, software development kits, and training frameworks...
read Sep 23, 2025Google launches MCP Server to democratize AI data access
Google launched the Model Context Protocol Server to provide developers with standardized access to public data from its Data Commons knowledge graph without requiring complex API integrations. The server builds on Anthropic's open MCP standard and aims to reduce AI hallucinations by giving large language models access to trusted public datasets, potentially democratizing data access for AI development at an unprecedented scale. What you should know: The MCP Server simplifies how AI agents consume publicly available data by eliminating the need for developers to navigate complex APIs. Data Commons provides public datasets from trusted sources for AI developers, data scientists...
read Sep 22, 2025Meta offers free Llama AI models to federal agencies through General Services Administration deal
Meta is providing its Llama artificial intelligence models to federal agencies at no cost through a new deal with the General Services Administration. The open-source nature of these models allows government agencies to maintain full control over data processing and storage while delivering significant value to taxpayers through reduced implementation costs. What you should know: The GSA announced Monday that Meta will offer its open-source AI models and tools to federal agencies for free, emphasizing cost savings and data control benefits. Federal agencies can build, deploy, and scale AI applications at lower costs since the models are publicly available. The...
read Sep 18, 2025Trump DOJ targets AI companies for anticompetitive, control-of-pipeline practices
The Trump administration is actively monitoring the artificial intelligence sector for anticompetitive behavior as part of its strategy to maintain U.S. AI dominance, according to Department of Justice officials. This enforcement approach signals that federal antitrust regulators view competition protection as essential for fostering innovation in the rapidly evolving AI landscape. What you should know: The DOJ is specifically targeting exclusionary practices that could limit access to critical AI infrastructure and resources. Assistant Attorney General Gail Slater emphasized that "the competitive dynamics of each layer of the AI stack and how they interrelate, with a particular eye towards exclusionary behavior...
read Sep 17, 2025DeepSeek’s $294K AI model becomes first to pass peer review
DeepSeek's AI model R1 has become the first major large language model to undergo peer review, with researchers publishing details in Nature revealing the reasoning-focused system cost just $294,000 to train. The landmark study provides unprecedented transparency into how the Chinese startup created a model that rivals OpenAI's offerings at a fraction of the cost, potentially reshaping expectations around AI development expenses and accessibility. What you should know: The peer-reviewed paper confirms DeepSeek's innovative approach to creating powerful AI without relying on competitor outputs. R1 excels at reasoning tasks like mathematics and coding, competing directly with US-developed models while costing...
read Sep 12, 2025Retrofuturistic engineer connects 2002 GameCube to modern AI for real-time dialogue
Software engineer Joshua Fonseca has successfully connected the 2002 GameCube classic Animal Crossing to modern AI language models, creating a mod that replaces the game's original dialogue with AI-generated conversations. The technical achievement bridges a 22-year gap between Nintendo's pre-internet console and cloud-based AI systems without modifying any game code, demonstrating how creative hacking can breathe new life into retro gaming experiences. How it works: Fonseca's Python script monitors game memory through the Dolphin emulator and communicates with AI models like GPT-4 or Gemini to generate real-time dialogue. The mod uses a "memory mailbox" technique, writing directly to specific GameCube...
read Sep 9, 2025UAE releases 32B-parameter AI model that matches OpenAI performance
The United Arab Emirates has released K2 Think, an open-source AI reasoning model that matches the performance of much larger systems from OpenAI and DeepSeek while using just 32 billion parameters compared to their 200+ billion. Developed by researchers at Mohamed bin Zayed University of Artificial Intelligence in Abu Dhabi and backed by tech conglomerate G42, this represents one of the strongest indicators yet that the UAE's massive AI investments are yielding competitive results in the global race for AI supremacy. What you should know: K2 Think is specifically designed for advanced reasoning tasks rather than being a complete large...
read Sep 3, 2025MIT’s FlowER AI predicts chemical reactions while conserving physics
MIT researchers have developed FlowER (Flow matching for Electron Redistribution), a new generative AI system that predicts chemical reactions while maintaining fundamental physical constraints like conservation of mass and electrons. The breakthrough addresses a critical limitation in existing AI models that often violate basic scientific principles by creating or destroying atoms during reaction predictions, potentially revolutionizing drug discovery and materials science. The big picture: Previous AI attempts at reaction prediction have struggled because they ignore fundamental physics, leading to scientifically impossible outcomes where atoms appear or disappear during reactions. Large language models like ChatGPT use computational "tokens" to represent atoms,...
read Sep 2, 2025Switzerland launches Apertus, world’s first open-source national AI model
Switzerland has launched Apertus, an open-source national Large Language Model developed by public institutions including EPFL, ETH Zurich, and the Swiss National Supercomputing Centre. The move positions Switzerland as the first country to offer a fully transparent AI model as public infrastructure, potentially setting a new standard for how nations approach AI development and data sovereignty. What you should know: Apertus represents a completely open approach to AI development, with full transparency across its entire training process. Users can inspect any part of the model's training, access comprehensive documentation, source code, and the datasets used to build it. The model...
read Sep 1, 2025Chile develops Latam-GPT, a 50B-parameter AI model for Latin America
The Chilean National Center for Artificial Intelligence is developing Latam-GPT, an open-source large language model specifically designed for Latin America and trained on regional languages and contexts. The project aims to help the region achieve technological independence by creating AI that understands local dialects, cultural nuances, and historical contexts that global models often overlook. What you should know: Latam-GPT represents a collaborative effort across Latin America to build regionally-focused AI capabilities. The model contains 50 billion parameters, making it comparable to GPT-3.5 in scale and complexity. It's trained on over 8 terabytes of text data from 20 Latin American countries...
read Aug 27, 2025Job alert: Open-source LiteLLM raises $1.6M, now hiring founding engineer
LiteLLM, an open-source LLM Gateway with 27,000+ GitHub stars, is seeking a founding backend engineer to help scale their platform that allows companies to call 100+ LLM APIs in OpenAI format. The Y Combinator-backed startup has raised $1.6M in seed funding and serves enterprise clients including NASA, Adobe, and Rocket Money who need unified access to multiple AI models. What you should know: LiteLLM provides both a Python SDK and FastAPI server that standardizes calls across major AI providers like OpenAI, Anthropic, Bedrock, Azure, and VertexAI. The role offers $160K-$220K salary plus 0.50%-3.00% equity for candidates with 1+ years of...
read Aug 25, 2025xAI open-sources Grok 2.5 with limits on AI model training
xAI has released Grok 2.5 as an open-source model, allowing developers to download, run, and modify the AI system through Hugging Face, a popular platform for sharing AI models. CEO Elon Musk announced that the upcoming Grok 3 will also go open source within six months, marking a significant shift toward accessibility that contrasts sharply with OpenAI's more restrictive approach to model distribution. What you should know: The open-source release comes with specific limitations designed to protect xAI's competitive interests. Users can download and tweak Grok 2.5's source code, but xAI's license prohibits using it to train, create, or improve...
read Aug 22, 2025Salesforce study shows GPT-5 fails over half of enterprise AI tasks
Salesforce AI Research has released MCP-Universe, an open-source benchmark revealing that even advanced AI models like OpenAI's GPT-5 fail more than half of real-world enterprise orchestration tasks. The benchmark tests how large language models interact with Model Context Protocol (MCP) servers—a system that lets AI models connect with external tools and data sources—across six enterprise domains, exposing significant limitations in current AI capabilities for business applications. What you should know: MCP-Universe evaluates AI models on practical enterprise tasks rather than isolated performance metrics, providing a more realistic assessment of AI readiness for business deployment. The benchmark tests models across six...
read Aug 21, 2025Chan Zuckerberg Initiative launches rBio AI model for virtual cellular biology simulations
The Chan Zuckerberg Initiative launched rBio on Thursday, the first AI model trained to reason about cellular biology using virtual simulations rather than expensive laboratory experiments. This breakthrough could dramatically accelerate biomedical research by allowing scientists to test biological hypotheses computationally before committing resources to costly lab work, potentially flipping the traditional paradigm where 90% of biology research happens experimentally. The big picture: rBio addresses a fundamental challenge in applying AI to biological research by creating the first conversational AI system that can answer complex biological questions in plain English while being grounded in rigorous scientific data. Traditional biological foundation...
read