The archive · 11 labs · 3,050 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.

1001

OpenAI Announces Partnership with Indeed to Integrate AI into Job Search

OpenAI announced a collaboration with Indeed to embed its AI technology into the job search experience, aiming to streamline candidate matching and improve hiring efficiency.

1002

Qwen3-Max-Thinking Release Notes

Qwen has released Qwen3-Max-Thinking, a flagship reasoning model featuring adaptive tool-use and a multi-round test-time scaling strategy to compete with GPT-5.2-Thinking and Claude-Opus-4.5.

1003

OpenAI Codex CLI Agent Loop Technical Overview

OpenAI explains the agent loop architecture of the Codex CLI, detailing how it manages model inference, prompt construction, and context window optimization via prompt caching and compaction.

1004

Ollama launch command release

Ollama has introduced the 'ollama launch' command in version 0.15+, allowing users to set up and run coding tools like Claude Code and OpenCode without manual environment variables or configuration files.

1005

Scaling PostgreSQL to power 800 million ChatGPT users

OpenAI scaled PostgreSQL to support 800 million users and millions of queries per second by utilizing a single-primary architecture with nearly 50 read replicas and aggressive workload optimization.

1006

ChatGPT Enterprise Adoption Patterns and GPT-5 Integration

OpenAI reports that ChatGPT is being adopted bottom-up across industries, with GPT-5 introducing a real-time router to automate the use of advanced features for business workflows.

1007

Praktika AI Language Learning Architecture

Praktika utilizes a multi-agent system powered by GPT-5.2 and GPT-5 Pro to provide personalized, adaptive language tutoring through real-time conversational AI.

1008

Anthropic Claude Constitution 2026 Release

Anthropic released a new, publicly licensed constitution for its Claude model, detailing the values, safety priorities, and training role that guide Claude’s behavior.

1009

Qwen3-TTS Release Notes: Open-Source Voice Design, Cloning, and Generation

Qwen has open-sourced the Qwen3-TTS family, featuring 0.6B and 1.7B models that enable high-fidelity voice cloning, natural language-based voice design, and ultra-low latency streaming speech generation across 10 languages.

1010

Debugging a Memory Leak in vLLM

Mistral AI resolved a system memory leak in vLLM caused by UCX's mmap hooking mechanism during disaggregated serving, fixing it by disabling the hooks or limiting the registration cache.

1011

Higgsfield AI Video Platform Integration with OpenAI Models

Higgsfield is a generative media platform that uses GPT-4.1, GPT-5, and Sora 2 to transform product links and simple ideas into cinematic, social-first short-form videos.

1012

AssetOpsBench: Bridging the Gap Between AI Agent Benchmarks and Industrial Reality

Hugging Face and IBM Research introduce AssetOpsBench, a comprehensive benchmark designed to evaluate multi-agent AI systems in industrial Asset Lifecycle Management, focusing on safety-critical failure modes and operational constraints.

1013

OpenAI Ending the Capability Overhang Report and Initiatives

OpenAI has released research on the AI capability overhang and expanded its OpenAI for Countries initiative to help governments integrate AI into education, health, and public services to prevent economic divergence.

1014

OpenAI Education for Countries program launch

OpenAI launched Education for Countries, a program that partners with governments and universities to embed AI tools like ChatGPT Edu and GPT‑5.2 into K‑12 and higher‑education systems, aiming to close the capability overhang and prepare students for an AI‑driven workforce.

1015

Designing AI-Resistant Technical Evaluations

Anthropic details how it evolved its performance engineering take-home test through three iterations to maintain evaluation signal as Claude models increasingly outperformed human candidates.

1016

Anthropic and Teach For All AI Training Initiative

Anthropic has partnered with Teach For All to launch the AI Literacy & Creator Collective, providing AI fluency training and Claude access to over 100,000 educators across 63 countries.

1017

OpenAI Horizon 1000: Advancing AI for Primary Healthcare

OpenAI and the Gates Foundation have launched Horizon 1000, a $50 million initiative to integrate AI capabilities into 1,000 primary healthcare clinics across African countries, starting in Rwanda, by 2028.

1018

OpenAI Stargate Community program announcement

OpenAI announced its Stargate Community program, committing to fund its own energy, minimize water use, and create local jobs as it expands AI infrastructure toward a 10 GW goal by 2029.

1019

Anthropic appoints Mariano-Florentino Cuéllar to Long-Term Benefit Trust

Anthropic announced that Mariano-Florentino (Tino) Cuéllar has joined its Long-Term Benefit Trust, bringing legal and international‑affairs expertise to guide the company’s public‑benefit mission as AI reshapes global governance.

1020

One Year Since the DeepSeek Moment: The Rise of China's Open Source AI Ecosystem

Hugging Face analyzes how the release of DeepSeek R1 catalyzed a massive shift toward open-source AI in China, leading to a surge in high-performance open models that now dominate global download metrics.

1021

ServiceNow and OpenAI Partnership for Actionable Enterprise AI

ServiceNow has partnered with OpenAI to integrate GPT-5.2 and multimodal capabilities into its AI Platform, enabling enterprises to automate complex workflows across IT, finance, and HR.

1022

Differential Transformer V2 Source Unavailable

The source content for Differential Transformer V2 is unavailable due to a 429 Too Many Requests error from the Hugging Face website.

1023

OpenAI Age Prediction for ChatGPT

OpenAI is deploying an age prediction model on ChatGPT consumer plans to automatically apply teen safety safeguards for users likely under 18.

1024

Waypoint-1: Real-time Interactive Video Diffusion from Overworld

Overworld has released Waypoint-1, a real-time interactive video diffusion model trained on 10,000 hours of game footage that allows zero-latency world interaction via text, mouse, and keyboard.

1025

Ollama adds experimental image generation with Z-Image Turbo and FLUX.2 Klein

Ollama announced experimental macOS image generation using the 6B‑parameter Z‑Image Turbo model and Black Forest Labs’ FLUX.2 Klein models, marking the first native text‑to‑image capability in the Ollama ecosystem.

1026

Anthropic Assistant Axis research

Anthropic introduced the “Assistant Axis,” a neural‑activation direction that identifies and stabilizes the Assistant persona in LLMs, and showed that activation capping along this axis reduces harmful persona drift while preserving capabilities.

1027

OpenAI Business Model and Compute Scaling Strategy

OpenAI has implemented a business model that scales with the value of intelligence, driving a 10x revenue increase from 2023 to 2025 through a diversified compute portfolio and multi-tier monetization.

1028

D4RT: Dynamic 4D Reconstruction and Tracking

Google DeepMind introduces D4RT, a unified AI model for 4D scene reconstruction and tracking that is up to 300x more efficient than previous methods.

1029

ChatGPT Go Global Release

OpenAI has expanded ChatGPT Go, a low-cost subscription plan priced at $8 per month in the US, to all available markets to increase global access to GPT-5.2 Instant.

1030

OpenAI Advertising Strategy and ChatGPT Go Expansion

OpenAI is expanding the $8/month ChatGPT Go tier to the U.S. and all available markets and will begin testing ads in the free and Go tiers to increase AI accessibility.

1031

Anthropic Claude Sonnet 4.5 Demonstrates Autonomous Multi‑Stage Attacks Using Only Open‑Source Tools

Anthropic announced that Claude Sonnet 4.5 can autonomously carry out multi‑stage attacks on realistic cyber ranges—including a full Equifax‑style data exfiltration—using only standard Bash commands on Kali Linux, highlighting rapidly lowering barriers for AI‑driven cyber operations.

1032

Anthropic Appoints Irina Ghose as Managing Director of India

Anthropic has appointed Irina Ghose as Managing Director of India to lead the company's expansion and the opening of its first office in Bengaluru.

1033

Ollama v0.14.0: Anthropic API Compatibility and Claude Code Support

Ollama v0.14.0 introduces compatibility with the Anthropic Messages API, enabling the use of Claude Code and other Anthropic SDK applications with local and cloud open-source models.

1034

OpenAI Investment in Merge Labs

OpenAI is participating in the seed round for Merge Labs, a research lab developing high-bandwidth brain-computer interfaces (BCIs) to bridge biological and artificial intelligence.

1035

OpenAI Initiative to Strengthen U.S. AI Supply Chain via Domestic Manufacturing

OpenAI has launched a Request for Proposals (RFP) to identify and enable U.S.-based manufacturing for critical AI infrastructure components, including data center hardware, consumer electronics, and robotics.

1036

Hugging Face Open Responses Inference Standard

Hugging Face has introduced Open Responses, an open inference standard based on the Responses API designed to replace the Chat Completion format for agentic AI workflows.

1037

Anthropic Economic Index January 2026 Report – Economic Primitives

Anthropic’s January 2026 Economic Index report introduces five new “economic primitives” that quantify how Claude is used, revealing geographic concentration, task complexity, skill bias, and productivity implications.

1038

Anthropic Economic Index January 2026 Report

Anthropic introduces economic primitives to measure AI's impact on labor productivity, finding that Claude provides the greatest speedups for complex tasks requiring higher education.

1039

OpenAI Codex with Ollama Integration

Ollama now enables the OpenAI Codex CLI to use open-weight models, allowing for local code execution and modification using models like gpt-oss:20b and gpt-oss:120b.

1040

Anthropic Claude for Life Sciences accelerates research with new tools and partnerships

Anthropic announced that Claude for Life Sciences, powered by the Opus 4.5 model, is being used in systems like Biomni, MozzareLLM, and Lundberg Lab’s hypothesis generator to dramatically speed up data analysis, experiment design, and gene‑target selection, marking a shift toward AI‑driven scientific discovery.

1041

OpenAI and Cerebras Partnership for Low-Latency Inference

OpenAI is partnering with Cerebras to integrate purpose-built AI systems into its inference stack to accelerate AI model responses and enable real-time AI interactions.

1042

Finding bugs with Claude and property-based testing

Anthropic researchers developed an AI agent using Claude that autonomously identifies bugs in large software projects by inferring code properties and applying property-based testing via Hypothesis.

1043

Veo 3.1 Release Notes: Enhanced Ingredients to Video and 4K Upscaling

Google DeepMind has released Veo 3.1, introducing improved identity and object consistency for image-to-video generation, native vertical video support, and upscaling to 1080p and 4K resolutions.

1044

Zenken Case Study: Scaling Sales and Operations with ChatGPT Enterprise

Zenken has integrated ChatGPT Enterprise to automate knowledge work, resulting in 50 million yen in annual outsourcing savings and a 30% increase in final proposal approval rates.

1045

Anthropic Labs: Incubating Experimental AI Products

Anthropic has expanded Labs, a dedicated team focused on incubating experimental products at the frontier of Claude's capabilities to accelerate the transition from research preview to scaled product.

1046

Anthropic launches Claude for Healthcare and expands Claude for Life Sciences with new connectors and Agent Skills

Anthropic announced Claude for Healthcare—HIPAA‑ready tools for providers, payers, and health‑tech firms—and added dozens of new scientific connectors and Agent Skills to Claude for Life Sciences, powered by the Opus 4.5 model.

1047

Information-Driven Design of Imaging Systems

BAIR researchers have developed a framework to evaluate and optimize imaging systems using mutual information, enabling high-performance hardware design without the need for task-specific decoders.

1048

OpenAI and SoftBank Group Partnership with SB Energy

OpenAI and SoftBank Group have each invested $500 million in SB Energy to accelerate the development of AI data center campuses and energy infrastructure, including a 1.2 GW site in Milam County.

1049

Datadog Integrates OpenAI Codex for System-Level Code Review

Datadog has deployed OpenAI Codex to over 1,000 engineers to identify systemic risks and cross-service dependencies in code reviews, successfully detecting 22% of historical incidents that human reviewers missed.

1050

Anthropic Constitutional Classifiers++ Release

Anthropic has introduced Constitutional Classifiers++, a two-stage ensemble defense system that reduces compute overhead to ~1% while eliminating known universal jailbreaks.