Securing top talent with our guide is a click away

Deel’s simplified a whole planet’s worth of information. It’s time you got your hands on our international compliance handbook where you’ll learn about:

• Attracting global talent

• Labor laws to consider when hiring

• Processing international payroll on time

• Staying compliant with employment & tax laws abroad

With 150+ countries right at your fingertips, growing your team with Deel is easier than ever.

Get the guide

Today:

• Anthropic, Palantir, AWS Collaborate on Defense

• Ex-OpenAI CTO Launches Fresh AI Team

• Greg Brockman Returns to OpenAI

• AI Startup Writer Hits $1.9B

• Baidu Adds No-Code App Builder

• Adobe Stock Introduces AI Customization Tools

Anthropic, Palantir, Amazon team up on defense AI

Anthropic, creator of the AI model Claude, has partnered with Palantir and Amazon Web Services (AWS) to bring advanced AI into U.S. defense and intelligence. Claude is designed with "constitutional AI," a built-in ethical framework, ensuring safe use in high-stakes environments. 

With Palantir’s data integration abilities and AWS's secure infrastructure, Claude can quickly analyze vast intelligence data—like satellite imagery and communications—helping human analysts detect potential threats. This collaboration aims to boost national security without replacing human judgment. 

WATCH THE VIDEO ON YOUTUBE

Ex-OpenAI CTO Launches Fresh AI Team

Former OpenAI CTO Mira Murati has started a new venture, attracting key former OpenAI team members, including research program manager Mianna Chen and researchers Barret Zoph and Luke Metz. This adds to a wave of talent loss at OpenAI, as notable departures also include safety systems head Lilian Weng. 

These exits mark a shift in OpenAI’s team composition, especially among its safety and post-training groups. Meanwhile, OpenAI aims to launch the full version of its o1 reasoning model by year-end, hoping it advances the company's AI capabilities amid these staffing changes.

Greg Brockman Returns to OpenAI

Greg Brockman, co-founder of OpenAI, has returned to the AI startup following a three-month leave from his role as president. Brockman is now collaborating with CEO Sam Altman on a new role focused on tackling significant technical challenges. 

This return follows a series of notable departures from OpenAI, including former CTO Mira Murati and head of safety systems, Lilian Weng. OpenAI has yet to comment on these developments.

AI Startup Writer Hits $1.9B

AI startup Writer has secured a $1.9 billion valuation, raising $200 million to expand its large language models beyond content creation. Led by CEO May Habib, Writer’s models—optimized for accuracy in business-specific tasks—are now used by companies like Uber, Qualcomm, and Dropbox for functions ranging from customer support to legal management. 

Writer aims to streamline enterprise workflows by developing AI agents that autonomously handle tasks across platforms. Competing with OpenAI and others, Writer focuses on cost-effective, tailored solutions for enterprises, positioning itself as a productivity “super app” that simplifies and automates various applications within companies.

Baidu Adds No-Code App Builder

At the Baidu World Conference, Baidu unveiled new AI tools, including I-RAG, a text-to-image generation technology that addresses image inaccuracies and a no-code app builder, Miaoda. Baidu’s CEO, Robin Li, emphasized the company’s focus on using its Ernie platform to handle tasks like text generation and coding without requiring coding expertise from users. 

Ernie now processes 1.5 billion daily interactions, up from 200 million in May. Additionally, Baidu introduced AI-powered smart glasses with a built-in assistant for photos, videos, and voice support. Unlike some competitors, Baidu isn’t pursuing a single AI "super app" approach.

Adobe Stock Introduces AI Customization Tools

Adobe has introduced AI tools in its stock photography service, allowing users to modify stock images with AI while compensating original creators. These tools let customers start with Adobe’s stock images and adapt them to specific needs without replacing traditional stock or creative contributors. 

According to Adobe’s VP Matthew Smith, this enhancement gives creators more income opportunities by leveraging AI for customization rather than generating entirely new images. Adobe’s approach ensures commercially safe content and legal compliance, setting it apart from competitors by prioritizing artist compensation and minimizing risks associated with AI-generated images.

🧠RESEARCH

Add-it: Training-Free Object Insertion in Images With Pretrained Diffusion Models

Add-it introduces a new approach for adding objects into images using text instructions without retraining. It extends diffusion models' attention mechanisms, balancing natural placement and scene preservation. With strong results on benchmarks and human preference in over 80% of tests, Add-it outperforms other methods without fine-tuning.

OmniEdit: Building Image Editing Generalist Models Through Specialist Supervision

OmniEdit is a versatile model designed to perform seven different image editing tasks on any image aspect ratio. By training with guidance from specialized models and using advanced sampling to enhance data quality, OmniEdit addresses limitations in previous methods. Human and automated tests show OmniEdit outperforms existing models significantly in editing success and adaptability.

Chinese SimpleQA: A Chinese Factuality Evaluation for Large Language Models

Chinese SimpleQA introduces the first detailed Chinese benchmark to assess the factual accuracy of large language models in answering short questions. Covering six major topics and 99 subtopics, this benchmark offers high-quality, static answers, ensuring reliable evaluation. Chinese SimpleQA aids developers in improving factuality for Chinese-language models.

M-Longdoc: A Benchmark For Multimodal Super-Long Document Understanding And A Retrieval-Aware Tuning Framework

M-LongDoc introduces a benchmark and retrieval-aware tuning method for understanding super-long, multimodal documents, which are challenging for humans to analyze due to their complexity. The benchmark includes recent, lengthy documents and open-ended question answering. M-LongDoc's tuning approach improves model accuracy by 4.6%, streamlining document analysis across text, figures, and tables.

Edify Image: High-Quality Image Generation with Pixel Space Laplacian Diffusion Models

Edify Image is a powerful diffusion model designed for photorealistic image generation with high pixel accuracy. Using a unique Laplacian diffusion method, it processes different frequency bands for refined detail. Edify Image supports various applications, including text-to-image generation, 4K upscaling, 360 HDR panoramas, and custom image finetuning.

🛠️TOP TOOLS

Rezi - Free AI Resume Builder.

Context Autopilot - AI productivity suite that learns like you, thinks like you, and uses tools like you.

Runway Gen-3 Alpha - A new frontier for fast, high-fidelity, controllable video generation.

Lamatic - A managed PaaS with a low-code visual builder, VectorDB, and integrations to apps and models for building, testing, and deploying high-performance GenAI apps on edge.

Vizology - Generate limitless business ideas, gain insights on markets and competitors, and automate business plan creation

📲SOCIAL MEDIA

🗞️MORE NEWS

• Amazon has committed $110 million to support AI research at universities through its Build on Trainium program, providing resources and training for advanced AI development. This initiative offers funding, access to high-performance Trainium chips, and encourages open-source research to foster innovation in machine learning and generative AI.

• Microsoft-backed startup Fastino has launched task-optimized AI models that run efficiently on CPUs instead of costly GPUs, reducing enterprise costs. These models, focused on specific tasks, offer high accuracy and are suitable for data-sensitive industries needing on-premise solutions.

• screengrasp is a new AI tool that controls computers more effectively than others. It handles high-quality images, supports multiple languages, understands entire screens, and works on various systems. By combining several AI models, it offers superior performance and ease of use.

• Unitree_IL_lerobot is an open-source project that helps train and test data from Unitree robots. It provides code and easy-to-follow instructions for setting up environments, controlling robots, collecting data, and conducting experiments, simplifying robotics research.

• Nous Research has launched the Forge Reasoning API Beta and Nous Chat. Forge enhances AI models, including Hermes 3, with advanced reasoning via Monte Carlo Tree Search, Chain of Code, and Mixture of Agents, boosting real-world problem-solving.

• Jasper has introduced Brand IQ and Marketing IQ tools to enhance control over AI-generated content for marketers, embedding brand guidelines and optimizing for marketing goals like SEO. These updates streamline content creation, helping brands maintain