DEV Community

Jay Suresh
Jay Suresh

Posted on Originally published at thenewsbay.com

Meta Orion, Grok 3, and ByteDance Doubao: The Week AI Really Changed

The last 24 hours produced three technology stories that, taken together, signal something bigger than any one announcement. Here's a clear-eyed look at what happened and why it matters.


Meta Finally Shows the World What AR Glasses Can Be

For years, augmented reality glasses have been stuck in a loop: impressive demos, awkward hardware, prices that kept them squarely in enthusiast territory. Meta changed the conversation on Monday with the consumer version of its Orion glasses — $1,299, launching spring 2027, with a 70-degree field of view that no competing device has matched.

What makes this different from previous attempts isn't any single spec. It's the form factor. Orion weighs 98 grams and looks like a heavy pair of sunglasses. That matters more than raw horsepower, because the biggest barrier to AR adoption has never been what the glasses can do — it's whether people are willing to wear them in public. Meta has built something you might actually wear.

The glasses run Meta AI natively, can translate conversations in real time, and are controlled by voice and a wrist EMG band rather than controllers. Pre-orders open September 15.

Read the full story on The News Bay →


Grok 3 Claims the Top of the AI Leaderboard — With Caveats

xAI released Grok 3 on Monday and, according to benchmarks the company published, it outperforms GPT-5, Gemini Ultra 3, and Claude 4 Opus on reasoning, math, and code. On MATH-500, Grok 3 scored 98.1%. On GPQA Diamond, a graduate-level science reasoning test, it hit 87.4%.

Are these numbers real? Probably mostly yes, with important caveats. Independent researchers who looked at the published methodology said the results are plausible — but also pointed out what they always point out when a lab releases its own benchmarks: the tests a company chooses to publish are the ones its model does best on. The true measure of whether Grok 3 is the best available model will come from independent evaluations over the next few weeks as researchers put it through its paces.

For developers, the more practical question is: how does it perform on your specific use case? The answer to that, as always, is to try it. Grok 3 is available now to X Premium subscribers and through the xAI API.

Read the full story on The News Bay →


ByteDance's Doubao Pro Just Became the World's Most Downloaded App

ByteDance launched its Doubao Pro AI assistant globally on Monday — 143 countries, 47 languages — and within 48 hours it topped the charts on both the App Store and Google Play, beating ChatGPT, Claude, and Gemini everywhere.

The product itself is strong: text, image, and video generation, document analysis, and a memory system that maintains context across conversations over weeks. It's free to download with a $9.99/month premium tier. ByteDance has stored non-Chinese user data outside China and built in additional safety filters for Western regulatory environments — moves that clearly anticipate the scrutiny that's already arriving from Washington.

Three U.S. senators have called for an emergency national security review. The European Commission is watching closely. Whether Doubao Pro's extraordinary launch week translates into sustained presence in Western markets will depend as much on regulators as on users.

Read the full story on The News Bay →


Three launches in one day, each of them genuinely significant. The AI race is no longer about who has the best model in a lab — it's about who puts the best product in the most hands.

For more tech coverage, visit The News Bay.

Top comments (0)