Researchers developed a deep reinforcement learning-based Transformer method to tackle the open shop scheduling problem, a complex challenge in industrial and service settings. The approach, trained on small-scale benchmark instances, demonstrated strong scalability by producing feasible schedules for much larger problems—often within 15% of best-known solutions—while outperforming several classical dispatching heuristics. The model offers a learning-driven alternative to traditional methods, requiring minimal input data to generate competitive results.

Researchers developed a Transformer-based deep reinforcement learning method to tackle the open shop scheduling problem, a difficult optimization task in industrial and service operations. Trained on smaller benchmark instances, the model produced feasible schedules and generalized to much larger problems, performing competitively with established dispatching heuristics and outperforming some simpler rules.

OpenAI has launched the Partner Network, a $150 million initiative aimed at accelerating enterprise AI adoption worldwide. The program will support global partners in deploying and transforming AI solutions for businesses.

OpenAI has launched the Partner Network, a new initiative aimed at helping global partners accelerate enterprise AI adoption, deployment, and transformation. The company is investing $150 million in the effort to support broader use of AI across businesses.

NVIDIA Nemotron 3 Ultra is now available on Ollama’s cloud as a 550-billion-parameter open model with 55 billion active parameters. NVIDIA says the model is designed for long-running agentic workflows, supports integrations with tools such as Claude, Hermes and OpenClaw, and offers strong accuracy, throughput and cost efficiency compared with other leading open models.

NVIDIA's Nemotron 3 Ultra, a 550-billion-parameter open model optimized for long-running, agentic workflows, is now available on Ollama’s cloud platform. The model delivers leading accuracy and throughput while reducing costs by up to 30% compared to other open models. It supports fast, affordable performance across hundreds of tool calls for tasks like coding, instruction following, and long-context processing.

Google DeepMind introduced Gemma 4 12B, a mid-sized, encoder-free multimodal model designed to run locally on consumer laptops with 16GB of RAM. The model supports native audio inputs, targets agentic multimodal tasks, and is positioned between the smaller E4B and the larger 26B Mixture of Experts model with a reduced memory footprint.

Google DeepMind unveiled Gemma 4 12B, a compact yet powerful multimodal AI model designed to run efficiently on laptops without separate encoders for audio and visual inputs. The model delivers near-top-tier performance with a smaller memory footprint, enabling advanced reasoning and agentic capabilities on consumer hardware. It marks a step forward in making high-performance AI more accessible for developers and everyday devices.

Julia Evans shares her writing approach, explaining she tailors her work for a specific audience—often her past self or a close friend—to make it more relatable and effective. The quote, highlighted by Simon Willison, reflects her strategy for clearer communication.

Simon Willison shared a Julia Evans quote about writing with a specific reader in mind rather than a broad audience. Evans describes often imagining “me, but 3 years ago” or a good friend as the person she is writing for.

Mistral announced a slate of AI Now Summit initiatives, including physics AI tools for industrial engineering, expanded partnerships with Airbus, BMW Group and ASML, and the acquisition of Emmi to strengthen scientific modeling capabilities. The company also introduced an upgraded Vibe agent for research, productivity and coding tasks, and plans a new 10 MW inference facility in Les Ulis, France, scheduled to open in Q3 2026.

The AI Now Summit 2026 showcased Mistral’s advancements in AI for industrial and enterprise applications, including partnerships with Airbus, BMW, and ASML to integrate custom AI models into design, simulation, and manufacturing. New tools like Vibe, an AI agent for coding and workflow automation, and a 10 MW inference facility in Les Ulis were also announced to enhance productivity and secure infrastructure. These developments aim to accelerate innovation in aerospace, automotive, and semiconductor industries while maintaining data control and security.