✷ 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.
vLLM Session-Aware Agentic Routing (SAAR) Release
vLLM introduces Session-Aware Agentic Routing (SAAR), a model selection policy that preserves session continuity and reduces costs for long-horizon LLM agents by preventing unsafe model switches during tool loops and provider-state transitions.
vLLM-Omni Accelerates Inference with AutoRound Quantization
vLLM-Omni now integrates Intel's AutoRound post-training quantization, enabling W4A16 quantization that cuts model size up to 62% while preserving accuracy and unlocking performance gains on Intel XPU and NVIDIA GPUs.
Anthropic Expanding Project Glasswing
Anthropic is expanding Project Glasswing to 150 new organizations to secure critical infrastructure using Claude Mythos Preview, aiming to establish new cybersecurity norms before the widespread release of powerful cyber-capable AI models.
OpenAI Policy and Political Advocacy Positions
OpenAI has clarified that it does not fund political action committees or candidates and maintains that AI policy should be shaped by a broad coalition of stakeholders rather than any single organization.
JetBrains Mellum2 Release
JetBrains has announced the release of Mellum2, a 12B Mixture-of-Experts (MoE) model, though the source material provided is an error page and contains no technical details.
Beyond LLMs: Why Scalable Enterprise AI Adoption Depends on Agent Logic – Summary
Hugging Face’s post on “Beyond LLMs: Why Scalable Enterprise AI Adoption Depends on Agent Logic” could not be retrieved, so no technical details can be summarized.
OpenAI Stargate: The Barn Data Center in Michigan
OpenAI has broken ground on The Barn, a 1GW data center campus in Saline, Michigan, as part of its Stargate program to expand AI infrastructure and support American reindustrialization.
OpenAI Frontier Models and Codex Now Available on AWS
OpenAI has integrated its frontier models and Codex into AWS, allowing enterprises to deploy OpenAI capabilities using existing AWS security, compliance, and billing workflows.
Qwen3.7-Plus release notes / what's new
Qwen3.7-Plus is a multimodal agent model that unifies vision and language to operate across GUI and CLI environments for complex software engineering and productivity automation.
vLLM on the DGX Spark: Architecture, Configuration, and Local Evaluation
vLLM provides a high-performance local inference endpoint for NVIDIA DGX Spark, enabling the deployment of large NVFP4 models like Nemotron-3-Super using a unified-memory architecture and OpenAI-compatible API.
OpenAI Disrupts 'Data Center Bandwagon' Influence Operation
OpenAI banned a cluster of China-linked accounts that used ChatGPT to generate social media content and operational plans to influence US audiences regarding AI power demand and harass Chinese dissidents.
OpenAI Disrupts PRC-Linked 'Tech and Tariffs' Influence Operation
OpenAI banned a cluster of ChatGPT accounts used by PRC-linked actors to generate pro-PRC narratives, criticize US tech policy, and attempt to discredit OpenAI through false claims of data compromise.
xAI Composer 2.5 Release
xAI has released Composer 2.5, a fast, state-of-the-art coding model integrated into Grok Build designed for long-running tasks and complex instruction following.
Anthropic Confidentially Submits Draft S-1 for Proposed IPO
Anthropic, PBC has confidentially submitted a draft registration statement on Form S-1 to the U.S. Securities and Exchange Commission for a proposed initial public offering of its common stock.
Braintrust Integration of OpenAI Codex
Braintrust uses OpenAI Codex to transform customer feature requests into working code previews in minutes, significantly accelerating their development feedback loop.
Boston Children's Hospital AI Integration and Rare Disease Diagnosis
Boston Children's Hospital has integrated an enterprise AI layer to optimize operations and diagnose over 40 previously unresolved rare conditions.
Qwen-VLA: Unifying Vision-Language-Action Modeling for Embodied Intelligence
Qwen-VLA is a general-purpose Vision-Language-Action model that unifies robotic manipulation, vision-language navigation, and cross-embodiment control into a single generalist policy model.
OpenAI Rosalind Biodefense and GPT-Rosalind Expansion
OpenAI has launched the Rosalind Biodefense initiative and expanded trusted access to GPT-Rosalind for government and allied partners to accelerate the development of biodefense and pandemic preparedness tools.
Profiling in PyTorch: A Beginner's Guide to torch.profiler
Hugging Face provides a comprehensive guide to using torch.profiler to identify bottlenecks, understand the CPU-GPU dispatch chain, and analyze the impact of torch.compile on kernel execution.
OpenAI shares a playbook for trustworthy third‑party evaluations of frontier AI models
OpenAI published a shared playbook that explains how third‑party evaluators should choose harnesses, elicit capabilities, and check validity to produce trustworthy assessments of frontier AI models.
xAI Grok Build 0.1 API Release
xAI has released grok-build-0.1, a coding model optimized for agentic tasks such as web development and debugging, available via the xAI API in public beta.
Mistral AI Now Summit 2026 Announcements
Mistral AI introduced a specialized AI stack for industrial engineering, the Vibe agent for long-horizon productivity, and a new 10 MW data center in Les Ulis, France.
Mistral AI Vibe Release Notes
Mistral AI has launched Vibe, a unified AI agent that integrates work and coding workflows across web, mobile, IDE, and CLI interfaces.
Endava's Agentic Organization Strategy with OpenAI Codex
Endava has transitioned into an agentic organization by using OpenAI Codex to codify senior architectural expertise and compress software delivery lifecycles from weeks to days.
Mistral AI Search Toolkit Public Preview
Mistral AI has released Search Toolkit in public preview, an open-source composable framework designed to unify ingestion, retrieval, and evaluation into a single production search pipeline for AI applications.
Speculators v0.5.0 release notes / what's new
Speculators v0.5.0 introduces DFlash algorithm support for single-pass draft token generation, unified online and offline training via vLLM's native hidden states extraction, and updated documentation.
vLLM Semantic Router Multimodal Routing and Vision Encoder Hardening
vLLM has introduced multimodal routing to the Semantic Router (VSR), enabling the system to use visual evidence as a first-class signal for request-level policy decisions while resolving critical implementation drifts between Rust/Candle and PyTorch paths.
Laguna XS.2 Inference Optimization with vLLM, Speculators, and LLM Compressor
Poolside and Red Hat AI have optimized the Laguna XS.2 33B-A3B MoE model for agentic coding tasks using vLLM integration, DFlash speculative decoding, and LLM Compressor quantization.
vLLM Native RL APIs Release
vLLM has introduced native weight syncing APIs and improved asynchronous RL support to standardize weight transfer between training and inference and eliminate deadlocks in large-scale DPEP deployments.
OpenAI Frontier Governance Framework
OpenAI has introduced the Frontier Governance Framework to align its safety and security practices with emerging legal requirements like the EU AI Act and California’s Transparency in Frontier AI Act.
Anthropic Series H Funding and Valuation Update
Anthropic has raised $65 billion in Series H funding, valuing the company at $965 billion post-money to expand compute capacity and accelerate AI safety research.
OpenJarvis 1.0 Release: Local-First Personal AI via Ollama
OpenJarvis 1.0 is an open-source framework for building local-first personal AI agents that integrates with Ollama to run models on personal hardware.
Mistral AI Physics AI Announcement
Mistral AI has integrated Emmi AI to launch Physics AI, a class of data-driven models that accelerate industrial engineering by predicting physical behavior in seconds rather than hours or weeks.
Mistral AI Physics AI Research and Emmi AI Acquisition
Mistral AI is advancing industrial engineering through the acquisition of Emmi AI and the development of foundational Physics AI models for aerospace, automotive, semiconductors, and energy sectors.
Cisco and OpenAI Codex Enterprise Integration
Cisco integrated OpenAI's Codex into its production engineering workflows, reducing feature development time from quarters to weeks and achieving a 10-15x increase in defect resolution throughput.
Building self-improving tax agents with Codex
OpenAI and Thrive Holdings developed Tax AI, a self-improving agent for Crete accountants that uses a Codex-driven loop to automate complex tax returns with up to 97% accuracy.
Reachy Mini Local Speech Backend Integration
Hugging Face has released a local speech-to-speech pipeline for Reachy Mini, allowing the robot to handle conversations fully locally using a cascaded VAD, STT, LLM, and TTS stack.
Delta Weight Sync in TRL Enables Trillion-Parameter Model Training with Minimal Bandwidth
Hugging Face announced Delta Weight Sync in TRL, a feature that reduces weight synchronization bandwidth in async RL training by over 100x by transmitting only sparse weight changes via Hugging Face Buckets, enabling disaggregated training without shared clusters.
OpenAI Election Information and Safeguards in 2026
OpenAI has announced a comprehensive set of safeguards for the 2026 election cycle, focusing on reliable information surfacing, cyber infrastructure defense, content provenance, and the prevention of model bias.
Warp Open Agentic Development and Oz Orchestration Platform
Warp is implementing Open Agentic Development using GPT-5.5 and its Oz orchestration platform to shift software engineering toward human-supervised agent fleets.
Anthropic Opens Milan Office to Support Italian Enterprise and Research
Anthropic has opened a new office in Milan, its sixth European location, to support Italian enterprises, developers, and research in the responsible scaling of Claude AI.
Grok Integration in Kilo Code
xAI has integrated Grok models, including Grok Build, into the Kilo Code agentic engineering platform for users with X Premium+ or SuperGrok subscriptions.
Coding Agents in the Social Sciences
A survey of 1,260 social scientists reveals that while 81% use AI chatbots, only 20% have adopted autonomous coding agents, with significant adoption gaps across gender, career stage, and institutional prestige.
EAGLE 3.1 Release Notes: Enhancing Speculative Decoding Robustness and Efficiency
EAGLE 3.1 introduces architectural improvements to solve attention drift, doubling acceptance length in long-context workloads and significantly increasing throughput via vLLM and TorchSpec integration.
Anthropic Appoints KiYoung Choi as Representative Director of Korea
Anthropic has appointed KiYoung Choi as Representative Director of Korea to lead go-to-market strategies and support the high adoption rate of Claude.ai in the South Korean market.
Hugging Face AI Agent Glossary: Defining Harness, Scaffold, and Agent Architecture
Hugging Face provides a standardized vocabulary for AI agents, defining the agent as the combination of a model, a harness for execution, and scaffolding for behavior definition.
OpenAI, Grupo Folha, and Grupo UOL Strategic Content Partnership
OpenAI has partnered with Brazil's Grupo Folha and Grupo UOL to integrate high-quality Brazilian journalism into ChatGPT, providing 900 million weekly active users with grounded, attributed reporting.
OpenAI Analysis: AI as a First Hire for Small Businesses
OpenAI reports that four million US users utilized ChatGPT in March 2026 to support small business operations, reducing administrative burdens and lowering the fixed costs of entrepreneurship.
Anthropic How We Contain Claude Across Products – Engineering Overview
Anthropic announced its multi‑layer containment strategy for Claude agents, detailing environment sandboxing, human‑in‑the‑loop controls, and egress filtering to cap blast radius across Claude.ai, Claude Code, and Claude Cowork.
Anthropic Co-Founder Chris Olah on Pope Leo XIV's Magnifica Humanitas Encyclical
Anthropic co-founder Chris Olah emphasizes the need for external moral and philosophical oversight from religious and civil society leaders to balance the commercial and geopolitical incentives of AI development.