You thought 2026 would start with a hangover and a whimper?

Nope.

We aren't even a full week into the new year and the AI world is already screaming at full volume.

Meta just panic-bought a startup for billions. OpenAI is reportedly eyeing a massive social media acquisition to fix their cash problem.

And Grok?

Well, Grok is currently explaining to governments why it’s generating... that.

If you’re still mentally on holiday break, we don't blame you. But you missed a lot.

Here's what matters.

Meta finally decided to stop getting beaten in the agent game.

According to The Wall Street Journal, Meta has agreed to acquire Manus, a Singapore-based agent startup, in a deal valued at more than $2 billion.

Manus was literally in the middle of raising new funding when Meta swooped in.

The signal is clear: Meta is done trying to build everything internally. They are buying their way to the front of the line.

Manus is famous for being a "general use" super agent. It can generate deep research reports and build custom websites with almost zero human input. It uses models from firms like Anthropic and Alibaba to get it done.

Interest in the company exploded after a demo in March, right around the time DeepSeek was making waves.

Now? Manus CEO Zhao "Red" Hong will report directly to Meta COO Javier Olivan.

What it means: Meta admits it has a gap in its AI portfolio.

They haven't released a major LLM since Llama 4 in April 2025.

By acquiring Manus, they instantly compete with Microsoft and Google in the "autonomous agent" space without spending another year in the lab.

Wait. OpenAI buying... Pinterest?

Holllllld up.

This sounds fake, but The Information reports that OpenAI has serious interest in acquiring the digital scrapbooking giant.

Why?

Two words: Images and Ads.

Pinterest has over 600 million users and a goldmine of image data. OpenAI wants to expand its online shopping and advertising capabilities to become the "everything app" for its billion weekly users.

Pinterest’s new "Performance Plus" ad suite is the real prize here. It saw a 24% higher conversion lift for advertisers last year.

Let's be honest.

OpenAI is amazing at chat. But they are terrible at selling ads compared to Google and Meta.

Buying Pinterest would let them leapfrog five to ten years of ad tech development overnight.

What it means: OpenAI is aggressively hunting for sustainable revenue beyond subscriptions.

They dominate the consumer side, but they need a real ad engine to justify their valuation.

This deal would instantly make them a top-tier player in the digital ad market, right alongside the giants they are trying to kill.

The people have voted. And Google is like DJ Khaled cuz all it does is win, win win.

LMArena released their end-of-year rankings for 2025.

Google Gemini 3 Pro absolutely cleaned up.

It ranked as the top overall frontier model of 2025 and became the first AI model to ever hit an ELO score of 1500.

Gemini 3 Pro took the top spot in general text, reasoning, and coding. It even beat out its own little brother, Gemini 3 Flash, which took second place.

Other heavy hitters like Grok 4.1 and Claude Opus 4.5 trailed behind.

But there was one upset.

In the text-to-image category, GPT-1.5 Image stole the crown. But it barely edged out Google’s Nano Banana Pro, which went AI viral like a hundred times in 2025. 

What it means: Google has successfully clawed its way back to the top of the AI food chain.

After a rocky start a few years ago, Gemini 3 Pro is now the standard for reasoning and coding.

If you are still ignoring Gemini in favor of older models, you are officially behind the curve.

CES kicks off today. And it is going to be messy. Like… spilling and oozing AI everywhere it shouldn't even be. 

But if you care about actual AI and not just shoving LLMs in every new toaster, then you might be happy to know Jensen Huang is delivering the CES keynote address today. 

We expect news on Blackwell Ultra chips and maybe some foundation model updates.

But the rest of the show?

It’s going to be a parade of companies shoving AI into things that do not need AI.

We are talking AI shoe. AI mops. And yes, probably AI water bottles that yell at you to hydrate.

Marketers are desperate for headlines, so expect "AI-powered" to be slapped on everything from shoes to refrigerators and see a ton of press coverage. 

The one bright spot? Smart glasses.

Meta is finally getting some real competition. A bunch of startups are unveiling AI-powered glasses this week that might actually be useful.

What it means: The "AI inside" marketing trend has officially hit peak stupidity.

Most of these products are gimmicks.

However, the flood of smart glasses suggests that wearable AI is the next legitimate hardware battleground, finally moving beyond the failed "pin" devices of last year.

This is disturbing. And gross.

Elon Musk’s AI chatbot Grok is under intense scrutiny this week.

Users generated explicit images of children on the platform, including images of a 14-year-old actress from Stranger Things.

The controversy is huge because Grok is currently authorized for official US government use.

Grok (the chatbot, not the company) posted on X that there were "isolated cases" where safeguards failed. They claim they are fixing it.

But governments aren't waiting.

Officials in India and France have launched investigations. France even referred the matter to the agency enforcing the EU’s Digital Services Act.

India’s IT minister gave xAI exactly 24 hours to explain how this happened.

What it means: The "anti-woke" lack of guardrails has finally hit a legal wall.

Governments will not tolerate platforms that generate CSAM.

This controversy could cost xAI their lucrative government contracts and force them to implement the strict safety filters they have spent years mocking.

We have an

AI product generating a billion dollars a year. And it is building itself.

Boris Cherny, the creator of Claude Code at Anthropic, confirmed the news this past week.

In the last 30 days? 100% of his code contributions were written by the AI.

He didn't write a single line manually.

The AI submitted almost 260 pull requests. It added 40,000 lines of code. It made 497 commits.

Cherny says his role has shifted from "coder" to "architect and verifier."

Claude Code now operates as an autonomous agent in the terminal. And with the new browser version, even non-coders can use it for complex workflows.

What it means: Software engineering has fundamentally changed.

We are no longer writing syntax. We are managing agents that write syntax.

If the creator of the tool isn't even writing code anymore, the "human in the loop" is rapidly becoming a "human on the side."

You read that right.

A pen.

According to The Information, OpenAI is actively developing a consumer AI device codenamed "Gumdrop."

Reports say one of the main concepts is a smart pen.

The device would combine handwriting capture with voice recording and always-on ChatGPT assistance.

Sam Altman reportedly wants a device that feels like a "cabin by the lake."

He wants a calm, low-distraction alternative to the smartphone. A "third device" that doesn't bombard you with notifications.

They are also looking at a wearable pin or a smart speaker. But the pen seems to be the frontrunner.

What it means: OpenAI knows it can't rely on Apple and Google forever.

They need their own hardware ecosystem to gather data and serve users directly.

A pen is a safer bet than a phone, avoiding a direct fight with the iPhone while capturing the one thing phones are bad at: handwriting.