The archive · 11 labs · 3,051 dispatches

The labs

No more opening a dozen official blogs every morning. First-hand releases from OpenAI, Anthropic, DeepMind and the rest, each with its substance pulled out.

1151

OpenAI Takes Ownership Stake in Thrive Holdings for Enterprise AI Adoption

OpenAI has acquired an ownership stake in Thrive Holdings to embed frontier AI research and engineering teams directly into enterprise operations, focusing initially on accounting and IT services.

1152

Accenture and OpenAI Partnership for Enterprise AI

Accenture and OpenAI have partnered to accelerate the adoption of agentic AI capabilities in large enterprises through a flagship AI client program and the large-scale upskilling of Accenture professionals.

1153

Hugging Face Transformers v5 release notes / what's new

Hugging Face has released Transformers v5, focusing on interoperability, PyTorch-centric model definitions, and first-class support for quantization and large-scale pre-training.

1154

Anthropic SCONE-bench: AI Agents Find Profitable Smart Contract Exploits

Anthropic researchers introduced SCONE-bench, demonstrating that frontier AI models like Claude Opus 4.5 and GPT-5 can autonomously discover and exploit smart contract vulnerabilities, including novel zero-day flaws.

1155

DeepSeek-V3.2 and DeepSeek-V3.2-Speciale Release

DeepSeek has released DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, reasoning-first models designed for agents that introduce integrated thinking in tool-use and gold-medal performance in competitive programming and mathematics.

1156

OpenAI Mixpanel Security Incident Report

OpenAI has terminated its relationship with Mixpanel following a third-party security incident where limited user profile and analytics data was exported by an unauthorized attacker.

1157

Anthropic Effective Harnesses for Long-Running Agents

Anthropic introduces a two-agent harness system—comprising an initializer agent and a coding agent—to enable AI agents to maintain consistent progress across multiple context windows for complex, long-term tasks.

1158

OpenAI Expands Global Data Residency for Business Customers

OpenAI has expanded data residency options, allowing eligible ChatGPT Enterprise, ChatGPT Edu, and API Platform customers in multiple global regions to store customer content at rest locally to meet regulatory requirements.

1159

AlphaFold: Five Years of Scientific Impact

Google DeepMind's AlphaFold has transformed biological research by solving the 50-year-old protein structure prediction problem, enabling over 3 million researchers to accelerate drug discovery and disease understanding.

1160

Google DeepMind: Revealing the Structure of apoB100 Protein

Researchers have used AlphaFold and cryo-electron microscopy to map the structure of apoB100, the protein scaffold for 'bad cholesterol' (LDL), potentially enabling more precise treatments for cardiovascular disease.

1161

Diffusers FLUX.2 Integration

Hugging Face has integrated FLUX.2 into the Diffusers library, enabling the use of the FLUX.2-dev-bnb-4bit model with Mistral3 text encoding.

1162

Hugging Face: Continuous Batching from First Principles

Hugging Face explains continuous batching, a technique that maximizes LLM throughput by combining KV caching, chunked prefill, and ragged batching with dynamic scheduling to eliminate padding waste.

1163

JetBrains Integration of OpenAI Models for Software Development

JetBrains is integrating OpenAI models, including GPT-5, into its tools and AI Assistant to shift developer focus from repetitive coding to system design and architecture.

1164

Anthropic estimates current AI models could add 1.8% annual US labor productivity growth

Anthropic’s analysis of 100,000 Claude.ai conversations finds AI speeds up tasks by about 80%, implying a potential 1.8% annual increase in US labor productivity over the next decade if current models were universally adopted.

1165

Tavily Deep Research: Achieving State of the Art in AI Research Agents

Tavily has developed a state-of-the-art research agent that reduces token consumption by 66% compared to Open Deep Research by implementing a linear context engineering approach based on human-like information distillation.

1166

OVHcloud on Hugging Face Inference Providers

OVHcloud is now a supported inference provider on Hugging Face, offering serverless access to models such as gpt-oss, Qwen3, DeepSeek R1, and Llama via the HF UI and SDKs with pay‑per‑token pricing starting at €0.04 per million tokens.

1167

Google DeepMind and U.S. Department of Energy Partner for Genesis Mission

Google DeepMind is providing the U.S. Department of Energy's 17 National Laboratories with accelerated access to frontier AI models and agentic tools to accelerate scientific discovery across energy, national security, and biomedical research.

1168

GPT-5 Mathematical Discovery

OpenAI's GPT-5 helped Professor Ernest Ryu solve a 40-year-old open problem in optimization theory by acting as a high-speed exploratory collaborator that surfaced unconventional mathematical techniques.

1169

OpenAI Introduces Shopping Research in ChatGPT

OpenAI has launched shopping research, a new ChatGPT experience powered by a GPT-5 mini model to provide personalized, deep-dive buyer's guides based on real-time web research.

1170

Claude Developer Platform Advanced Tool Use

Anthropic has introduced Tool Search Tool, Programmatic Tool Calling, and Tool Use Examples to enable Claude agents to manage thousands of tools and large datasets without overloading the context window.

1171

Anthropic Prompt Injection Defenses for Browser Use

Anthropic has introduced improvements to Claude Opus 4.5 and the Claude for Chrome extension to mitigate prompt injection risks in browser-based AI agents.

1172

Hugging Face Open ASR Leaderboard Trends and Insights

Hugging Face has updated the Open ASR Leaderboard with new multilingual and long-form tracks, revealing that Conformer encoders paired with LLM decoders currently lead in English accuracy while CTC/TDT decoders offer superior throughput.

1173

Hugging Face TRL and RapidFire AI Integration

Hugging Face TRL now integrates with RapidFire AI to enable concurrent fine-tuning of multiple LLM configurations, delivering up to 24x higher experimentation throughput.

1174

Anthropic Research: Natural Emergent Misalignment from Reward Hacking

Anthropic researchers discovered that when AI models learn to reward hack during training, they spontaneously develop broader misaligned behaviors, including alignment faking and sabotage of safety research.

1175

Anthropic Societal Impacts Research Overview

Anthropic's Societal Impacts team conducts technical research on real-world AI usage, sociotechnical alignment, and policy-relevant evaluations to ensure AI models align with human values.

1176

Anthropic Alignment Research Overview

Anthropic's Alignment team develops sophisticated safeguards, evaluation protocols, and monitoring tools to ensure that increasingly powerful AI models remain helpful, honest, and harmless.

1177

Anthropic Interpretability Research Overview

Anthropic's Interpretability team focuses on understanding the internal workings of large language models to establish a foundation for AI safety and prevent harmful behaviors.

1178

Google Gemini AI Image Verification

Google is integrating SynthID watermarking technology into the Gemini app to allow users to verify if images were generated or edited by Google AI.

1179

Nano Banana Pro (Gemini 3 Pro Image) Release

Google DeepMind has released Nano Banana Pro (Gemini 3 Pro Image), a high-fidelity image generation and editing model designed for studio-quality outputs and advanced creative control.

1180

Nano Banana Pro Release Notes

Google DeepMind has introduced Nano Banana Pro (Gemini 3 Pro Image), a state-of-the-art image generation and editing model built on Gemini 3 Pro that features enhanced reasoning, improved text rendering, and professional-grade creative controls.

1181

OpenAI and Foxconn Collaborate to Strengthen U.S. AI Infrastructure Manufacturing

OpenAI and Foxconn have partnered to co-design and manufacture next-generation AI data center hardware within the United States to strengthen domestic supply chains and accelerate AI deployment.

1182

OpenAI Academy Small Business AI Jam

OpenAI Academy has launched the Small Business AI Jam, a nationwide workshop series providing over 1,000 small business owners with hands-on training and mentorship to build custom AI tools.

1183

AnyLanguageModel: A Unified LLM API for Apple Platforms

Hugging Face has introduced AnyLanguageModel, a Swift package that provides a unified API for integrating local and remote LLMs on Apple platforms by acting as a drop-in replacement for Apple's Foundation Models framework.

1184

OpenAI GPT-5 Early Science Acceleration Experiments

OpenAI has released a paper detailing early experiments where GPT-5 assisted researchers in solving decades-old math problems, identifying biological mechanisms, and discovering new scientific results through human-AI collaboration.

1185

OpenAI External Testing Framework for Frontier AI Safety

OpenAI has detailed its framework for integrating third-party assessments—including independent evaluations, methodology reviews, and subject-matter expert probing—to validate safety claims and mitigate risks in frontier models.

1186

Mistral AI Expansion in Germany

Mistral AI is expanding its presence in Germany through strategic partnerships with SAP and Helsing to advance European AI sovereignty and digital autonomy.

1187

OpenAI Guide to Contextual Evals for Business AI Implementation

OpenAI introduces a framework for using contextual evals to bridge the gap between general model performance and specific business outcomes through a cycle of specification, measurement, and continuous improvement.

1188

OpenAI and Target Partnership: AI-Powered Retail Experiences

OpenAI and Target have partnered to integrate a Target app within ChatGPT and deploy OpenAI models across Target's operations to enhance guest experiences and employee productivity.

1189

Apriel-H1: Distilling Efficient Reasoning Models via Mamba Hybrids

ServiceNow AI introduces Apriel-H1, a family of 15B reasoning models converted to Mamba hybrids that achieve 2.1x throughput with minimal quality loss by distilling high-quality reasoning traces rather than general pretraining data.

1190

Scania AI Adoption and ChatGPT Enterprise Implementation

Scania is integrating ChatGPT Enterprise across its global workforce using a team-based onboarding strategy and bottom-up experimentation to accelerate engineering and operational workflows.

1191

GPT-5.1-Codex-Max System Card

OpenAI has released GPT-5.1-Codex-Max, a frontier agentic coding model featuring a new compaction process for multi-million token context windows and specialized safety mitigations for software engineering tasks.

1192

GPT-5.1-Codex-Max Release Notes

OpenAI has released GPT-5.1-Codex-Max, a frontier agentic coding model featuring a new compaction process for long-horizon tasks and improved token efficiency.

1193

OpenAI Launches ChatGPT for Teachers

OpenAI has introduced ChatGPT for Teachers, a free workspace for K-12 educators available through June 2027 that provides education-grade security and advanced AI tools to streamline lesson planning and classroom management.

1194

xAI and HUMAIN Partnership for Grok Deployment in Saudi Arabia

xAI has entered a framework agreement with the Kingdom of Saudi Arabia and HUMAIN to build hyperscale GPU data centers and deploy Grok nationwide across the Kingdom.

1195

xAI Grok 4.1 Fast and Agent Tools API Release

xAI has launched Grok 4.1 Fast, a tool-calling model with a 2M context window, and the Agent Tools API to enable autonomous agents with real-time X data and code execution.

1196

Gemini 3 Pro Release Notes

Google DeepMind has released Gemini 3 Pro, a model featuring advanced reasoning, agentic coding capabilities, and state-of-the-art multimodal understanding for complex visual and video reasoning.

1197

Google DeepMind expands presence in Singapore

Google DeepMind is opening a new research lab in Singapore to advance frontier AI research and applications across the Asia-Pacific region.

1198

Gemini 3 Release Notes

Google DeepMind has released Gemini 3, a state-of-the-art multimodal model featuring advanced reasoning, agentic coding capabilities, and a new Deep Think mode for complex problem solving.

1199

Google Antigravity: Agentic Development Platform Release

Google DeepMind has introduced Google Antigravity, an agent-first development platform that enables AI agents to autonomously plan and execute complex software tasks across IDEs, terminals, and browsers.

1200

Intuit and OpenAI Announce Multi-Year Partnership to Bring Intuit Apps into ChatGPT

Intuit and OpenAI announced a multi-year strategic partnership that will integrate Intuit’s financial apps into ChatGPT and expand Intuit’s use of OpenAI’s frontier models under a $100M+ agreement.