You found global talent. Deel’s here to help you onboard them

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:

• OpenAI Plans Full Code Automation

• Baidu Introduces ERNIE X1, 4.5

• OpenAI, Google Seek AI Training Rights

• DeepSeek AI Now Heavily Guarded

• Court Rejects Musk’s OpenAI Lawsuit

OpenAI "Full Code Automation" Coming This Year…

OpenAI’s CPO Kevin Wheel predicts that AI will surpass human programmers in coding by 2025, citing advances in reasoning-based AI models. OpenAI’s latest coding models have rapidly improved, with the upcoming versions expected to outperform top human coders. 

This shift could democratize software creation, enabling non-engineers to build complex applications. OpenAI also plans to monetize AI agents, projecting massive revenue growth from software development automation.

WATCH THE VIDEO ON YOUTUBE

Baidu Introduces ERNIE X1, 4.5

Baidu has introduced two new AI models, ERNIE X1 and ERNIE 4.5, to compete in China’s growing AI sector. ERNIE X1 boasts strong reasoning abilities and operates at half the cost of its rival DeepSeek R1. ERNIE 4.5 excels in multimodal understanding, language skills, and humor detection. Baidu aims to challenge global leaders like OpenAI but faces stiff competition in adoption.

Why This Matters:

1. Cost Efficiency: AI models delivering high performance at lower costs push affordability in AI development.

2. Advancements in Multimodal AI: ERNIE 4.5 enhances understanding across text, images, and audio, a crucial step toward more human-like AI.

3. Global AI Competition: China’s AI growth pressures U.S. firms, driving faster innovation worldwide.

OpenAI, Google Seek AI Training Rights

OpenAI and Google are urging the U.S. government to allow AI models to train on copyrighted content, arguing that restricting access could cause the U.S. to lose its AI edge to China. They claim fair use policies are crucial for innovation, while competitors like DeepSeek face fewer restrictions. The debate highlights growing legal and ethical concerns, as AI firms face lawsuits over unauthorized data use.

Why This Matters:

1. AI Copyright Battles: The outcome could reshape how AI companies access and use data.

2. National Security Implications: AI access to data is being framed as a competitive necessity against China.

3. Legal Precedents for AI Training: A decision favoring OpenAI and Google could set a global standard for AI copyright policies.

DeepSeek AI Now Heavily Guarded

DeepSeek, a rising AI startup in China, is now under tight control due to its national significance. Employees involved in AI research face travel restrictions, with some surrendering passports. The Zhejiang government is screening potential investors, limiting foreign influence. DeepSeek’s breakthrough AI model rivals OpenAI’s at lower costs, making it crucial to China’s tech ambitions. The company’s independence from outside funding adds to its strategic importance.

Why This Matters:

• It shows how government rules can affect the freedom and movement of AI experts.

• It highlights the impact of political and trade pressures on AI development.

• It emphasizes the growing role of startups in advancing and integrating AI technology.

🧠RESEARCH

Transformers without Normalization

Researchers show that Transformers, a type of AI model, can work well without normalization layers. They replace them with Dynamic Tanh (DyT), a simple mathematical function. DyT mimics how normalization behaves and achieves equal or better performance across various AI tasks. This challenges the belief that normalization is essential.

CoSTAast: Cost-Sensitive Toolpath Agent for Multi-turn Image Editing

CoSTA* is a method for efficient multi-step image editing. It combines AI planning with graph search to select cost-effective tools. CoSTA* balances cost and quality, adapts to failures, and switches between different AI tools. It outperforms existing models, improving accuracy and efficiency in complex image-editing tasks.

Charting and Navigating Hugging Face's Model Atlas

Researchers create an "atlas" to navigate Hugging Face’s vast model repository. By mapping documented models, they visualize trends and predict attributes like accuracy. To fill gaps, they use common training patterns to infer missing details. Their work improves model discovery and analysis, with an interactive tool for public use.

🛠️TOP TOOLS

Logo Redesign AI - AI-powered tool that transform and enhance your brand’s visual identity with ease and efficiency.

Copyleaks AI - Designed to identify and flag text generated by AI models.

Figurative Language Checker - Scans your text and ultimately identifies figurative language elements like similes, metaphors, and personification.

Neural Frames - Audioreactive AI animations for musicians, creatives and visual artists.

Augie AI - Designed to make video-first marketing accessible and intuitive for businesses of all sizes.

📲SOCIAL MEDIA

🗞️MORE NEWS

A court rejected Elon Musk’s attempt to hinder OpenAI, dismissing key claims. OpenAI argues Musk’s lawsuit is self-serving, aiming to control or compete with them. They reaffirm their commitment to a strong non-profit supported by a for-profit business.

A leaked Apple meeting reveals delays in Siri’s AI improvements, with no guarantee they’ll arrive in iOS 19. Leadership acknowledges internal frustrations, marketing missteps, and quality issues, emphasizing Apple won’t release unfinished features despite competitor pressure.

Google’s Gemini AI model can remove watermarks from images, raising copyright concerns. Users found it effective at erasing watermarks, though it struggles with complex ones. Critics argue this lacks proper restrictions, potentially violating copyright laws. Google hasn’t commented.

Google's Gemini models now understand videos, allowing users to analyze YouTube content via Google AI Studio. The system transcribes audio and examines video frames, enabling summaries, translations, and descriptions. Currently in preview, it supports up to eight hours daily.