✷ The archive · 11 labs · 3,414 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.
GPT-Red: Unlocking Self-Improvement for Robustness
OpenAI has introduced GPT-Red, an automated red-teaming model trained via self-play reinforcement learning to identify vulnerabilities and adversarially train future models, such as GPT-5.6 Sol, to be more robust against prompt injections.
Anthropic Commits $10 Million CAD to Canadian AI Research
Anthropic is investing $10 million CAD in Canadian research institutions and expanding API access for startups to support the next generation of responsible AI development.
Real World VoiceEQ: Measuring Human Quality in Voice AI
Hugging Face and Hume AI have introduced Real World VoiceEQ, a human-grounded benchmark designed to evaluate the emotional, acoustic, and conversational quality of voice AI beyond traditional technical metrics.
vLLM TML Inkling Support
vLLM has announced Day-0 support for TML Inkling, a 1T-parameter multimodal model, delivering up to 380 tok/s/user on 4 GB200 GPUs through specialized architectural optimizations.
Thinking Machines Inkling Release Notes / What's New
Thinking Machines has released Inkling, a 1 trillion parameter multimodal open model featuring a 1M context window and native support for image, text, and audio inputs.
Grok Build Open Source Release
xAI has open-sourced Grok Build, a coding agent and terminal user interface (TUI) that supports local-first execution and an extensible extension system.
OpenAI Guide to Managing AI Investments in the Agentic Era
OpenAI outlines a strategic framework for enterprise AI investment, shifting the focus from token pricing to useful work per dollar and outcome-based ROI.
vLLM and TileRT Integration for Latency-Critical Serving
vLLM has integrated TileRT 0.1.5 as a pluggable decode engine to provide native per-user decode speed for latency-critical workloads while maintaining vLLM's standard prefill and serving infrastructure.
How data science teams use ChatGPT Work
OpenAI has detailed how data science teams can use ChatGPT Work to transform raw data, dashboards, and business context into review-ready analysis assets.
How Sales Teams Use ChatGPT Work
OpenAI introduces ChatGPT Work and a dedicated sales plugin to help sales teams synthesize account context and customer data into actionable sales artifacts like pipeline briefs and account plans.
How Sales Teams Use ChatGPT Work
OpenAI introduces workflows for sales teams using ChatGPT Work and a dedicated sales plugin to synthesize CRM data and communication logs into actionable sales artifacts.
How Data Science Teams Use ChatGPT Work
OpenAI introduces ChatGPT Work as a tool for data science teams to synthesize fragmented data, dashboards, and notes into review-ready analysis assets.
How Canada Uses Claude: Insights from the Anthropic Economic Index
Canada ranks second globally in per capita Claude adoption among top users, with usage patterns driven primarily by workforce composition and official bilingualism policies.
Claude for Creative Work: New Connectors for Creative Software Integration
Anthropic has released a set of connectors that integrate Claude with industry-standard creative software, enabling natural language control, automated production tasks, and enhanced tool mastery for creative professionals.
Anthropic Opens Sydney Office and Appoints Theo Hourmouzis as GM for Australia and New Zealand
Anthropic has officially opened its Sydney office and appointed Theo Hourmouzis as General Manager for Australia and New Zealand to expand its regional presence and enterprise AI adoption.
Google DeepMind Launches ATL Saathi for Indian Educators
Google DeepMind has launched ATL Saathi, a Gemini-powered AI assistant designed to provide 24/7 planning and training support for educators in India's Atal Tinkering Labs.
EAGLE-3 Speculative Decoding on AMD Instinct GPUs
vLLM and AMD Quark have implemented an end-to-end pipeline for EAGLE-3 speculative decoding on AMD Instinct GPUs, achieving throughput speedups of up to 2.00x for Kimi-K2.5 and 1.79x for MiniMax-M2.5.
Anthropic Claude Values Across Models and Languages
Anthropic released a study showing that Claude’s expressed values can be reduced to four axes that differ systematically across model versions and the top 20 languages, revealing measurable shifts in deference, warmth, depth, and candor.
Deutsche Telekom AI-Native Transformation
Deutsche Telekom is transitioning to an AI-native telecommunications provider by redesigning its operating model, integrating AI into network operations and voice communications, and scaling ChatGPT Enterprise across its 200,000 employees.
Anthropic and UST Partnership for Physical AI and Enterprise Integration
Anthropic and UST are partnering to integrate Claude into physical AI engineering processes and enterprise systems across the semiconductor, automotive, healthcare, telecom, and banking sectors.
Profiling in PyTorch (Part 3): Attention is all you profile
Hugging Face’s Profiling in PyTorch (Part 3) shows how different attention implementations appear in PyTorch profiler traces, revealing performance trade‑offs of naive, in‑place, math, efficient, flash, and cuDNN backends.
vime ROCm Support for AMD Instinct GPUs
vLLM has announced ROCm support for vime, enabling end-to-end reinforcement learning post-training workflows to run natively on AMD Instinct MI300X and MI355X GPUs.
Getting Started with ChatGPT: Guide to Core Features and Workflows
OpenAI provides a foundational guide to ChatGPT, detailing the distinction between Chat and Work modes, prompt engineering basics, and the integration of voice capabilities for enhanced productivity.
Anthropic Golden Gate Claude Interpretability Demo
Anthropic demonstrated the ability to surgically alter a model's internal activations to create Golden Gate Claude, a version of Claude 3 Sonnet that is obsessed with the Golden Gate Bridge.
Anthropic Long-Term Benefit Trust Governance Structure
Anthropic has introduced the Long-Term Benefit Trust (LTBT), an independent body designed to ensure the company's governance aligns with the long-term benefit of humanity as AI capabilities advance.
Anthropic Appoints Ben Bernanke to Long-Term Benefit Trust
Anthropic has appointed former Federal Reserve Chair and Nobel laureate Dr. Ben Bernanke to its Long-Term Benefit Trust to provide economic expertise on the societal impacts of advanced AI.
Anthropic Launches Hard Questions Initiative to Address AI Societal Impact
Anthropic has launched a new initiative to solicit and publicly track responses to the public's most difficult questions regarding the societal, economic, and ethical implications of advanced AI.
GPT-5.6 Release: New Preferred Model for Microsoft 365 Copilot
OpenAI has introduced GPT-5.6 as the preferred model for Microsoft 365 Copilot, improving productivity across Word, Excel, PowerPoint, Chat, and Cowork through higher-quality outputs and better token efficiency.
Mistral Studio: Version Control for Prompts and Skills
Mistral AI has introduced a system of record in Studio to provide versioning, ownership, and traceability for prompts and skills, enabling faster iteration and governed production deployment.
OpenAI Bio Bounty Program Transition and GPT-5.6 Scope
OpenAI has transitioned the GPT-5.5 Bio Bug Bounty into the ongoing private OpenAI Bio Bounty Program, increasing the reward for universal jailbreaks to $50,000 for GPT-5.6 and GPT-5.5.
ChatGPT Work and GPT-5.6 Release
OpenAI has launched ChatGPT Work, an agentic system powered by GPT-5.6 that can execute multi-step workflows across apps, create documents and web apps, and perform scheduled tasks.
GPT-5.6 Sol, Terra, Luna: Performance, Features, and Availability
OpenAI launches GPT-5.6, introducing the Sol, Terra, and Luna models with improved performance per dollar, ultra multi-agent capability, and strengthened safeguards.
Ollama Funding and Open Model Platform Expansion
Ollama has raised $88M to expand its platform for running and scaling open AI models, serving 8.9 million developers and 85% of the Fortune 500.
Anthropic introduces Reflect with Claude beta
Anthropic has launched a beta feature called Reflect with Claude that allows users to track, visualize, and refine their AI usage patterns to better integrate AI into their daily lives.
Anthropic Claude Robotics Evaluation Shows Rapid Gains on High‑Level Control but Limited Direct Torque Mastery
Anthropic’s July 2026 study finds that newer Claude models can reliably supervise pretrained locomotion and manipulation policies, while direct low‑level motor control remains unreliable across robot bodies.
Anthropic Constitutional Classifiers: Defending against universal jailbreaks
Anthropic announced Constitutional Classifiers, a new input‑output filtering system that reduces universal jailbreak success to under 5% while adding only modest compute overhead and a 0.38% increase in harmless refusals.
OpenAI National Security Principles and Government Partnerships
OpenAI has published its National Security Principles to provide transparency into how the company handles government partnerships and the use of its technology in national security and law enforcement.
OpenAI Analysis of SWE-Bench Pro Coding Evaluation Flaws
OpenAI has retracted its recommendation for SWE-Bench Pro after an audit revealed that approximately 30% of the benchmark's tasks are broken due to issues like overly strict tests and underspecified prompts.
Robostral Navigate Release Notes
Mistral AI has introduced Robostral Navigate, an 8B model that enables robots to autonomously navigate complex environments using only a single RGB camera, achieving 76.6% success on unseen R2R-CE benchmarks.
OpenAI Academy AI Skills Jam for K–12 Educators
OpenAI Academy, in partnership with the Walton Family Foundation, is launching an AI Skills Jam to provide over 1,600 K–12 educators with hands-on training to integrate AI into teaching and administration.
Hugging Face Native-speed vLLM Transformers Modeling Backend
Hugging Face has updated the transformers vLLM backend to match or exceed the throughput of custom vLLM implementations by dynamically applying inference-specific layer fusions at runtime.
Introducing GPT‑Live‑1 and GPT‑Live‑1 mini: OpenAI’s New Full‑Duplex Voice Models
OpenAI announces GPT‑Live‑1 and GPT‑Live‑1 mini, full‑duplex voice models that enable natural, interruptible conversations while delegating complex reasoning to GPT‑5.5, rolling out globally to ChatGPT users.
Anthropic and AE Studio Introduce GRAM for Dual-Use Knowledge Control
Anthropic and AE Studio have developed Gradient-Routed Auxiliary Modules (GRAM), a method to isolate dual-use knowledge into removable compartments to prevent misuse without sacrificing general model performance.
Hugging Face and Amazon SageMaker Studio Integration
Hugging Face has introduced a deep-link integration with Amazon SageMaker AI, allowing developers to move from model discovery to fine-tuning or deployment in SageMaker Studio with a single click.
Hugging Face Models on Foundry Managed Compute
Microsoft Foundry now integrates a curated, weekly-refreshed catalog of Hugging Face open-weight models that can be deployed in one click onto Foundry Managed Compute for enterprise-grade operationalization.
Intelligence is Free, Now What? Data Systems for, of, and by Agents
BAIR researchers propose a new framework for data systems redesigned for agentic workloads, agentic state management, and agent-driven system synthesis as AI inference costs approach zero.
Hugging Face and SkyPilot Integration for Zero-Egress AI Storage
Hugging Face and SkyPilot have integrated to allow AI workloads to run on any cloud provider with zero-egress costs for reading models and datasets stored on the Hugging Face Hub.
MUFG AI-Native Transformation with OpenAI
MUFG is partnering with OpenAI to implement ChatGPT Enterprise for 35,000 employees and develop AI-driven customer experiences to transition into an AI-native financial group.
Australian Payments Plus Integration of ChatGPT Enterprise and Codex
Australian Payments Plus (AP+) has deployed ChatGPT Enterprise and Codex to accelerate technical investigations, streamline complex knowledge work, and reduce product prototyping time from weeks to a single day.
LeRobot v0.6.0 release notes / what's new
LeRobot v0.6.0 introduces world model policies, a expanded VLA model zoo, a unified reward models API, and a new deployment CLI with DAgger-style human-in-the-loop corrections.