Google's coding assistant is now free for everyone

Hello AI lovers! Here’s what you need to know about AI today:

👉 Google just made its coding AI assistant free for everyone

👉 Microsoft launched a multi-modal AI model for both digital and physical world

👉 OpenAI is rolling out Deep Research to all paying ChatGPT users

and many more!

📧 Did someone forward you this email? Subscribe here for free to get the latest AI news everyday!

Read time: 5 minutes

What’s going on: Google has announced that its Gemini Code Assist tool is now available for free to individual developers worldwide. This decision positions Gemini Code Assist as a formidable alternative to other popular tools like GitHub Copilot, Amazon Q Developer, and Anthropic’s Claude Code. Powered by the fine-tuned Gemini 2.0 model, the free version offers an impressive 180,000 code completions per month and 240 chat requests a day, far surpassing the 2,000 monthly completions provided by GitHub Copilot and Cursor AI free tier plans.

What does it mean: Once again Google is democratizing cutting-edge AI technologies for the global developer community. This tool stands out not only for its generous usage limits but also for its robust technical capabilities. It boasts a 128,000-token context window, enabling it to handle larger codebases effectively, which enhances its ability for debugging, generating code, and offering contextual suggestions. Supporting all programming languages in the public domain, the tool has been optimized for real-world coding scenarios.

More details: 

• Gemini Code Assist integrates seamlessly with widely used development environments/editors such as Visual Studio Code, JetBrains IDEs, Firebase, Android Studio, and GitHub, where it also provides AI-driven code reviews for both public and private repositories.

• Interested to learn more about the features of this brilliant tool? Read Google’s official blog post.

• Learn how to install Gemini Code Assist on VS Code or JetBrains IDEs.

What’s going on: Microsoft Research has introduced Magma, a new foundation model designed to power multimodal AI agents capable of operating seamlessly across both digital and physical environments. Magma represents a significant leap forward in AI development, aiming to create agents that can perceive, reason, and act in diverse contexts. Magma integrates multiple data types such as text, images, audio, and sensor inputs into a unified framework. This allows it to tackle complex tasks that require a holistic understanding of the world, bridging the gap between virtual interactions and real-world applications.

What does it mean: Magma has an ability to adapt to a wide range of scenarios, from assisting with digital workflows to navigating physical spaces. For instance, it could help users troubleshoot technical issues by analyzing visual and auditory cues or guide robots through dynamic environments using real-time sensory data. Microsoft emphasizes that Magma’s architecture is built for scalability and flexibility, enabling it to learn from vast, diverse datasets and improve over time.

More details:

• The adaptability is key to Magma’s potential, as it can be fine-tuned for specific industries like healthcare, manufacturing, or education.

• Magma is a Vision-Language-Action (VLA) model that leverages advanced machine learning techniques to combine perception, reasoning, and action into a cohesive loop, mimicking how humans navigate complex environments.

• Magma introduces a novel training paradigm centered on two key innovations: Set-of-Mark (SoM) and Trace-of-Mark (ToM) annotations. These techniques equip the model with a structured understanding of tasks in both user interface navigation and robotic manipulation domains.

• Interested in learning more about details and accessing the project’s code? Visit Magma’s official page.

📃 OpenAI is rolling out its "deep research" web browsing agent to all paying ChatGPT users, offering 10 queries per month for Plus, Team, Enterprise, and Edu subscribers, and increasing the limit to 120 queries for Pro users.

📽 Alibaba has made its AI video generation models, part of the Wan2.1 series, freely available globally. These models are accessible through Alibaba Cloud's ModelScope and Hugging Face.

💰 Meta is reportedly in talks to build a massive AI data center campus that could cost over $200 billion, with potential locations in Louisiana, Wyoming, and Texas. However, a Meta spokesperson has denied the claim, stating that any information beyond publicly disclosed plans is "pure speculation".

💼 Perfect, an Israeli AI startup, raised $23M to leverage AI in order to improve the efficiency and accuracy of recruitment processes.

📉 Anthropic's latest AI model, Claude 3.7 Sonnet, reportedly cost a few tens of millions of dollars to train, suggesting that training state-of-the-art models is becoming relatively cheaper compared to the hundreds of millions spent on models like GPT-4.

🐋 DeepSeek has reopened access to its API after a nearly three-week pause due to capacity constraints, allowing customers to top up credits again.

📱 Lovable, a Swedish AI startup that allows anyone to create functional apps using AI prompting, has raised $16 million reaching $17 million in ARR and 30,000 paying customers while enabling users to build over 25,000 new products daily.

🤖 Apptronik, a humanoid robot maker, has partnered with Jabil to pilot its Apollo robot in manufacturing tasks and eventually have Jabil manufacture the robot itself. This follows a previous pilot with Mercedes-Benz and positions Apptronik among competitors like Agility, Boston Dynamics, Figure, 1X and Tesla in the humanoid robotics industry.

AI + Data analytics

DigitalOcean - Staff AI/ML Engineer

Intuit Credit Karma - Staff Architect AI Applications

VISA - Director Software Engineering - Automation & AI Agent

Block - Deep Learning Research Intern (Summer 2025)

Thank you for staying with us like always! If you are not subscribed, subscribe here for free to get more of these emails in your inbox! Cheers!