Google releases its ‘most powerful model’ yet, OpenAI valuation hits $300 billion, is AI use making us become dumber?

Hello hello! We meet again.

While SignalGate might be the biggest AI scandal everyone’s talking about, let’s not forget that the tech industry moves fast, and some genuinely cool things are happening in the AI space as well. From Google’s latest AI brainpower flex to OpenAI’s image generator breaking the internet, here’s what’s been shaking up the AI world this week.

Google releases its ‘most powerful model’ yet

Google just dropped Gemini 2.5, its latest AI model that basically “pauses to think” before answering. The first in this lineup, Gemini 2.5 Pro Experimental, is now live for developers and premium users ($20/month). Google says all future models will have built-in reasoning, a feature that’s becoming the new battleground in AI.

Reasoning models, which OpenAI kicked off last year, are better at math, coding, and fact-checking but also cost more to run. Google’s been dabbling in this tech, but Gemini 2.5 is its boldest attempt yet to outdo OpenAI’s “o” series.  

Performance-wise, Gemini 2.5 Pro beats OpenAI, Anthropic, and DeepSeek on some coding tasks but lags behind Anthropic’s latest model in software development. It also scores pretty well on a massive knowledge test.  

One standout feature? A huge context window, 1 million tokens (about 750,000 words), with plans to double that soon. That’s more than the entire Lord of the Rings series in one go.  

No word yet on API pricing, but Google says details are coming soon.

OpenAI launches its most advanced AI image generator to date

OpenAI’s latest ChatGPT update is letting people create various kinds of visuals in seconds. It became so popular that the system crashed under demand. The trend kicked off with OpenAI and quickly spread, even reaching governments. 

The latest GPT-4o update can generate images and even videos in a bunch of styles, from South Park to classic claymation. However, one style dominated: Studio Ghibli. And that is going to raise a lot of copyright headaches for the company. More on that in our ‘AI Picture of The Month’ section.

ChatGPT and OpenAI’s Sora flooded social media with Ghibli-fied pop culture, politics, and memes—like The Sopranos and even Trump arguing with Zelensky.  

Of course, memes weren’t spared. Viral posts reimagined classics like “distracted boyfriend” and Ben Affleck smoking, while one standout showed Elon Musk playing with spoons (a nod to his recent dinner antics with Trump).  

Elon Musk’s xAI buys X

Elon Musk has officially merged xAI with X (formerly Twitter) in an all-stock deal, valuing X at $33 billion, which is down from the $44 billion investors recently paid. He’s framing it as a strategic move to integrate data, AI models, and distribution. 

But let’s be honest. X has been struggling with declining ad revenue, controversy, and investor uncertainty. By rolling X into xAI, Musk is essentially repackaging a messy social media platform as an exciting AI play, giving investors a shinier (and potentially more profitable) asset to hold onto. 

And since neither company is publicly traded, the details of the deal are pretty murky. No word on whether Musk himself got a financial boost from it.

Meanwhile, Musk is still dealing with an SEC lawsuit over his Twitter buyout, raising even more questions about the timing of all this. 

Is AI use making us become dumber?

AI makes work faster, but it might also be making us dumber. Or at least less critical in our thinking, according to a new study by Microsoft and Carnegie Mellon.

Researchers found that the more people relied on AI for tasks, the less they engaged in deep thinking. Instead of analyzing problems, they focused more on verifying AI responses and managing tasks. AI users also produced less diverse outcomes compared to those who worked independently. 

The big irony? By handling routine tasks, AI stops people from practicing their judgment, leaving them unprepared when complex issues arise. While AI boosts efficiency, overreliance on it could weaken problem-solving skills in the long run.  

With ChatGPT alone having 300 million monthly users, the impact could be huge. The good news? Researchers believe AI tools can be designed to encourage more critical thinking rather than replace it.

Apple faces first lawsuit over AI delay

Apple’s in hot water again—this time, over its much-hyped Apple Intelligence features. A new lawsuit, filed in a San Jose federal court, accuses Apple of false advertising, claiming customers bought AI-ready iPhones and other devices expecting features that never actually arrived.  

The lawsuit argues Apple’s ads led people to believe these AI tools would be available at launch, but instead, users got a half-baked or missing experience. Now, frustrated buyers are seeking damages in a class-action case.  

This legal mess adds to Apple’s AI struggles, with reports that CEO Tim Cook has lost faith in AI chief John Giannandrea’s ability to get things done.

You can read more about it here.

OpenAI valuation hits $300 billion

OpenAI just pulled in a massive $40B at a jaw-dropping $300B valuation, it announced in a blog post. SoftBank led the charge, with Microsoft and other usual suspects chipping in. 

A big chunk, about $18B, is going into OpenAI’s Stargate project, which is a network of AI data centers. The company says this cash will help push AI research, boost compute power, and keep ChatGPT running for its 500M weekly users.

If OpenAI were public, it’d be the 28th biggest company in the S&P 500, bigger than Chevron and almost as big as T-Mobile. Its valuation has nearly doubled since October’s $6.6B raise.

SoftBank is likely now OpenAI’s biggest backer, even surpassing Microsoft’s $13B investment. 

Did a friend forward this e-mail to you?

IE₊ SUPPORT INTERESTING ENGINEERING
_{Invest In Science And Engineering}

Enjoy exclusive access to the forefront of AI content, highlighting trends and news that shape the future. Join a community passionate about AI, delve into the latest AI breakthroughs, and be informed with our AI-focused weekly premium newsletters. With IE+, AI reporting goes beyond the ordinary - and it is Ad-Free.

AI PICTURE OF THE MONTH

OpenAI’s latest AI image generator has sparked a massive controversy—this time for its uncanny ability to produce Studio Ghibli-style artwork that looks nearly indistinguishable from the real thing.

Some users are loving it, amazed at how easily they can whip up Ghibli-esque masterpieces. But artists and fans? Not so much. They’re calling it outright theft, accusing OpenAI of profiting off stolen creativity. 

A 2016 video of Ghibli co-founder Hayao Miyazaki resurfaced, where he called AI-generated art an “insult to life itself.” Miyazaki, a die-hard hand-drawn animator, was “utterly disgusted” by AI-generated work, making this trend even more ironic.

Here’s the bad news for those hoping for legal action. Japan is one of the only major countries where it’s explicitly legal for AI models to train on copyrighted works. So even if OpenAI did use Ghibli images, it wouldn’t be illegal under Japanese law.

ChatGPT itself plays coy on whether Ghibli art was used for training but acknowledges that, given Japan’s relaxed stance, it could have happened.

AI TUTORIAL OF THE MONTH

We have a new image generator AI model, which is being touted as the best text-to-image model yet. Reve AI dropped Reve Image 1.0, which supposedly nails prompt accuracy, aesthetics, and even typography. 

Third-party testers rank it #1 in image quality, beating Midjourney v6.1, Google’s Imagen 3, and others. No word yet on API access, pricing, or whether it’ll be open source, but it’s off to a strong start.

What makes it special? Besides turning text into images, it lets you tweak existing ones with simple language commands like changing colors, adjusting text, and shifting perspectives. It’s also good at rendering text, making it a solid competitor to Ideogram for things like logos and branding.  

You can try it for free at preview.reve.art. No complex prompt skills are needed.  

Here is a video of a tutorial for Reve Image 1.0 to get you started: