The AI world just imploded. 

Microsoft agentized your live data. 

(Yeah, we made up a word.) 

Google silently dropped humanity's most powerful AI while everyone made anime selfies with an OpenAI image model that’s literally unbelievable.

All in ONE. SINGLE. WEEK.

And that’s not even half of the week’s biggest AI news. 

(Almost) Every Monday, we cut through the B.S. and bring you the AI News that Matters. 

Make sure to catch today’s episode before diving deeper into the recaps. 

Let's feast, shorties.

Microsoft just flipped the entire enterprise AI ecosystem upside down with major announcements to its Copilot Studio platform.

They’ve unveiled a new Deep Reasoning agent and Agent Flows inside of Copilot Studio.

Their deep reasoning capabilities tap directly into your dynamic Microsoft Graph data. No more static uploads going stale while you're making decisions.

The company revealed that over 400,000 AI agents were created in Copilot Studio just last quarter, showcasing adoption that's spreading faster than gossip in a small town.

Ray Smith, Microsoft's VP of AI agents, confirmed on Friday that agent flows launched give businesses unprecedented access to deterministic business logic with AI reasoning.

We had an exclusive interview with a Microsoft exec on Friday to bring you the inside scoop.

What it means:

Microsoft is eating everyone's enterprise lunch while we're distracted by pretty pictures.

Their agent ecosystem creates a moat no competitor can cross because of their Microsoft 365 integration.

The real AI revolution isn't generating anime selfies – it's transforming how actual businesses operate.

Best part?

It’s all low-code and natural language base.

SoftBank is reportedly throwing FORTY BILLION DOLLARS at OpenAI.

Billion with a B. Sheeeesh.

Bloomberg reported that OpenAI's next funding round could reach a staggering $40 billion, being led by SoftBank's multi-tiered investment.

The structure? Fascinating. SoftBank plans to lead with $7.5B initially, followed by $2.5B from other investors. Later this year, SoftBank could contribute another $22B, with others adding $7B+.

But there's a MASSIVE catch.

According to Reuters, OpenAI must finalize its transition from nonprofit to for-profit entity by the end of 2025 to secure the full amount.

Miss the deadline? SoftBank could slash $20 BILLION from the deal. Talk about pressure.

Previous valuation: $157B from October's $6.6B raise. New potential valuation: A mind-boggling $300B.

(Update: after our reporting this morning, this story was CONFIRMED this afternoon. See the updated news above in the AI News section.)

What it means:

Now we see why Elon's lawsuits are desperately trying to block OpenAI's conversion.

This isn't about ethics – it's about preventing a competitor from securing $40B. If OpenAI gets this cash, they'll have resources no other AI company can match.

Apple's cooking up an AI health revolution under the code name "Project Mulberry."

Bloomberg reports that Apple is advancing its healthcare technology with an AI-powered doctor and redesigned health app that will analyze data from devices like Apple Watch.

The system will provide tailored recommendations when it spots concerning patterns. Showing signs of high blood pressure? Here come personalized dietary suggestions.

Their Health+ app (unofficial name) will include food tracking for the first time, putting MyFitnessPal and Noom directly in Apple's crosshairs.

Reports indicate Apple is collaborating with in-house physicians and plans to expand with specialists for educational content – possibly even featuring a celebrity doctor for extra engagement.

What it means:

Apple isn't competing in general AI. They're weaponizing health data nobody else can access.

While everyone else builds chatbots, Apple has health sensors on millions of wrists.

Could this be the AI that Apple actually gets right?

Maybe.

Elon just bought Twitter. Again. Sort of.

According to Reuters, his AI company XAI purchased X (formerly Twitter) in a $45B all-stock deal including $12B in debt.

This values the social platform at $33B – TRIPLE recent estimates of just $10B and way closer to Musk's original $44B purchase price than anyone expected.

The acquisition positions Musk to merge resources between XAI and X, consolidating data, infrastructure, and talent to enhance Grok's training capabilities.

XAI's Memphis-based supercomputer cluster called "Colossus" is reportedly the largest in the world, designed to train next-generation models like Grok 3.

What it means:

No serious enterprise will touch Grok regardless of technical capabilities.The platform it's now integrated with consistently ranks worst for misinformation in multiple studies.

When AI's biggest challenge is trustworthiness, tying your model to Twitter is corporate no-no.

OpenAI quietly supercharged GPT-4o with massive improvements that barely anyone noticed.

The updated version shot from fifth to second place on LM Arena leaderboards, surpassing EVERY OTHER MODEL except Google's just-released Gemini 2.5 Pro.

Even more shocking? It now outperforms OpenAI's newer GPT-4.5 model in head-to-head human preference tests.

Users are calling it "unhinged" due to reduced restrictions, with OpenAI CEO Sam Altman highlighting improvements in coding, instruction-following, and creativity.

And in a victory for everyone's sanity: fewer default emojis. The emoji-pocalypse has been postponed.

What it means:

OpenAI aint done with GPT4o.

Even though we have GPT-4.5 and the ramblings of GPT-5, it’s clear that GPT4o is a powerhouse still that isn’t going away anytime soon.

A federal judge just greenlit The New York Times' copyright lawsuit against OpenAI, rejecting dismissal requests in a case that could redefine AI training.

Filed in December 2023, the lawsuit accuses OpenAI of siphoning millions of Times articles without permission or payment to train its GPT models.

Attorneys for The Times claim their content represents one of the largest sources of copyrighted text used to build ChatGPT, alleging the AI sometimes regurgitates articles verbatim.

The judge narrowed the scope slightly but allowed the primary copyright infringement claims to proceed, promising a detailed opinion soon.

A key point of contention? Market substitution – publishers fear chatbots summarizing news could divert readers from their websites, killing ad revenue.

What it means:

This case could reshape how all AI models are trained going forward.

If NYT wins, every major publisher will file similar suits within days.

OpenAI might need to retrain models without news content and pay significant damages.

The industry's future might hinge on copyright law, not technical capabilities.

Anthropic basically invented X-ray vision for AI brains.

Researchers at Anthropic created a new tool akin to an fMRI scan for artificial intelligence that reveals how large language models process information and make decisions.

Their Cross-Layer Transcoder (CLT) identifies circuits of neurons linked to specific reasoning tasks, offering unprecedented insights into AI's internal logic.

The bombshell discovery? Multilingual models like Claude don't think in separate languages at all. They use shared neural circuits to process universal concepts THEN translate the output to your requested language.

This breakthrough could improve model auditing, safety, and prevent hallucinations by making the "black box" of AI transparent for the first time.

What it means:

This is the most important AI research this year that nobody's talking about. Understanding how models actually reason is the key to building safe AGI.

These systems are developing internal representations beyond human language. While other labs hide behind black boxes, Anthropic is making AI transparent.

OpenAI unleashed GPT-4o image generation into the wild and broke the entire internet.

The company officially launched native image generation capabilities for ChatGPT users, marking a major milestone that instantly went mega-viral across social platforms.

Demand was so intense that OpenAI reported it was literally "melting their GPUs" due to what they called "biblical demand," forcing them to delay free user access and institute rate limits for paid accounts.

Unlike the older DALL-E models, this system integrates directly with the conversation interface, allowing users to refine images through natural language with unprecedented precision in text rendering, creative styles, and contextual understanding.

Sam Altman described it as a "new high watermark for creative freedom" as the entire internet frantically generated Studio Ghibli versions of their family photos.

What it means:

Creative agencies just experienced their Napster moment. $20/month now buys what previously required elite design teams.

No prompt engineering needed – just normal conversation produces professional results. Traditional design roles face massive disruption as the barrier between "pro" and "amateur" collapses.

Google just casually broke the AI universe while no one was looking.

They introduced Gemini 2.5 Pro featuring an unprecedented MILLION-token context window, enabling it to process extensive datasets including text, audio, images, video, and code repositories simultaneously.

The model immediately claimed the #1 spot on LM Arena by almost 40-point margin – the largest gap ever recorded when typical improvements are measured in single digits.

On "Humanity's Last Exam" benchmark, it scored 18% – double DeepSeek R1's 8.6% and comfortably ahead of OpenAI's o3 mini at 14%.

In a shocking weekend move, Google made this powerhouse available to free users immediately, with an upgrade to TWO MILLION tokens expected soon.

What it means:

Google just leapfrogged everyone while we were distracted by pretty pictures. This context window transforms what's possible with AI in practical business applications.

It’s….. SO GOOD!

Here’s a lil example of a siderunner game we one-shotted. (Chicago fans should love this!)