Grok 3 Is Now Free for All X Users

In partnership with

ChatGPT Turns Text Into PDF in Seconds

Presented By Mindstream

Get Your Free ChatGPT Productivity Bundle

Mindstream brings you 5 essential resources to master ChatGPT at work. This free bundle includes decision flowcharts, prompt templates, and our 2025 guide to AI productivity.

Our team of AI experts has packaged the most actionable ChatGPT hacks that are actually working for top marketers and founders. Save hours each week with these proven workflows.

It's completely free when you subscribe to our daily AI newsletter.

Subscribe to get instant access

AI & Tech News

#1. Gemini AI will soon turn any photo into a short video clip (Source)

Google is partnering with Honor to debut this groundbreaking Image-to-Video feature, powered by Gemini's Veo 2 model, launching with the Honor 400 series on May 22nd. Users will be able to animate any photo—whether a selfie or a famous painting—into a 5-second video without needing any text prompt. The AI handles it automatically inside the phone’s Gallery app.

This feature is free for the first 2 months for Honor 400 users, with a cap of 10 video generations per day. Eventually, it will require a Google subscription, likely Gemini Advanced. Currently, Veo 2 only supports text-to-video generation for subscribers, but image-to-video is being tested with select approved users at $0.50 per second.

#2. DeepSeek AI Is More Efficient (Source)

A new study by Greenly shows DeepSeek AI uses far fewer resources than rivals—just 2.78M GPU hours and 2,000 Nvidia chips to train, compared to Meta’s 30.8M hours and over 16,000 chips. It's also more energy-efficient than ChatGPT, which uses 25,000+ chips.

Forced by U.S. sanctions, DeepSeek optimized efficiency through a mixture-of-experts model, activating only needed computing power. As an open-weight model, it can run on local devices, reducing reliance on energy-hungry data centers.

#3. China Develops Fastest, Most Efficient Transistor—No Silicon Needed (Source)

Chinese researchers at Peking University have created a silicon-free transistor using a 2D material called bismuth oxyselenide.

Key Facts:

• 40% faster than Intel’s latest 3nm chip

• Uses 10% less power

• Based on a GAAFET (gate-all-around) design for better performance

• Outperforms chips from Intel, TSMC, and Samsung

• No need for silicon—uses new bismuth-based materials

• Can be made with existing chip tech

Scientists say it could replace silicon and solve size and power limits in future processors.

#4. Gemini’s Canvas Is Surprisingly Great for Writing and Coding (Source)

Gemini Canvas is an interactive space where you can:

• Write blog posts, study guides, reports, etc.

• Code and preview HTML or Python

• Build tools, visualizations, and even small games

• Co-create with AI and get real-time suggestions

#5. Obama Warns AI Will Take Over Most Coding Jobs (Source)

Former President Obama recently stated that AI is set to replace 60-70% of programming jobs, leaving only "the best coders" behind. Speaking at Hamilton College, Obama explained that AI is already automating many standard coding tasks, and soon, most workers in tech won't need to code at all.

While top-tier coders who integrate AI into their work will thrive, many routine programming jobs will disappear. Industry experts, like Nvidia CEO Jensen Huang, agree, emphasizing that AI is closing the technology gap and making coding less relevant.

Bill Gates also weighed in, predicting that AI will make many jobs, including medicine and teaching, more common and less specialized in the next decade. However, he believes some jobs, like watching baseball, will remain human-only.

AI's impact on coding jobs is already evident. In the past two years, over 27% of U.S. computer programming jobs have disappeared, largely due to AI automating routine tasks. While this shift continues, human coders will still play key roles in creative problem-solving and overseeing complex projects.

#6. Apple's AI-Powered Smart Glasses to Use New Model (Source)

Apple is working on AI-powered wearables, including smart glasses, to compete with Meta's Ray-Bans. Expected around 2027, these glasses will feature AI enhancements and be paired with AirPods that also have AI capabilities.

Apple's new FastVLM (Visual Language Model) is designed for fast, high-resolution image processing on devices. It’s up to 3.2 times faster and 3.6 times smaller than similar models, making it ideal for local AI processing without cloud dependence. FastVLM reduces processing time significantly, providing near-instant feedback with fewer resources needed.

#7. Elon Musk’s xAI Colossus Supercomputer Aims for 1 Million GPUs (Source)

Elon Musk’s xAI supercomputer in Memphis, TN, has 200,000 Nvidia GPUs, drawing 150MW of power from the local grid. The goal is to scale up to 1 million GPUs to compete with rivals like Oracle Cloud, which has 131,072 GPUs.

Launched in 2023, xAI’s Colossus supercluster grew quickly, with construction completed in just 122 days. In 92 days, the GPU count doubled from 100,000 to 200,000. Initially, xAI used natural gas generators for power, but now the Memphis grid provides the necessary 150MW, with 150MW in Tesla batteries to support operations.

Phase Two plans to expand to 300MW by fall 2025, as GPU numbers increase. Colossus is used to train large language models (LLMs) like Grok, Musk’s AI tool, and recently released Grok 3 beta.

Thanks for your time!

Shailesh & OpenAILearning team