China's Deepseek just dropped V3 on GitHub with a license permitting full commercial use, modification, and distribution. No restrictions. Zero gatekeeping.

Their mixture of expert (MOE) architecture plus V3 optimization is matching GPT-4o and Claude 3.5 in logical reasoning and problem-solving benchmarks. The accuracy stats don't lie.

You can pull this beast straight from Hugging Face without needing a supercomputer. Already optimized for coding and mathematics applications.

And API costs are bafflingly affordable. 

DeepSeek’s v3: 

• $0.14/million tokens input

• $0.28/million tokens output

GPT 4o mini: 

• $0.15/million tokens input

• $0.60/million tokens output. 

What it means: 

US tech giants are about to get bodied by Chinese open-source innovation in 2025. 

We’ve been talking about this for months. 

Keep an eye out on China’s LLMs in 2025. 

Meta's launching AI-generated accounts with full profiles, custom personalities, and independent posting abilities through their AI Studio platform.

Their current AI Studio already hosts "hundreds of thousands" of AI characters. Full rollout is planned across multiple countries over the next two years.

What it means: 

Zero details on content moderation, misinformation controls, or how they'll flag AI vs human accounts. Totally fine, right? 

These AI personas will interact with both human users and other AI accounts. Your aunt's political posts are about to get real interesting.

Facebook's speedrunning the death of authentic social connection. But hey, at least the AI drama will be entertaining.

According to reports, OpenAI's dropping the nonprofit angle and opting to become a Public Benefit Company (PBC) after securing $6.6B in funding earlier this year at a $157B valuation.

OpenAI just went from "we're saving humanity" to "we need that venture capital" real quick. They're transforming into a public benefit corporation after landing a $6.6B bag at a $157B valuation. And they're yanking those profit caps off for investors within two years.

OpenAI’s prolly gonna need TRILLIONS (of cash and compute) to chase AGI in the long run. Not billions - TRILLIONS. 

That nonprofit status looking real uncomfortable when you need GDP-level funding. And while Elon's out here throwing shade with (shaky) lawsuits, OpenAI's just trying to secure the bag.

The nonprofit arm's staying around like that one friend who still uses Facebook, but let's be real - this is all about that sweet, sweet capital.

What it means: 

Y'all really thought OpenAI was gonna build AGI with good nonprofit vibes and GitHub stars? Lolz. 

This is like being shocked that your favorite indie band signed with a major label. It’s inevitable to make it big. 

Those trillion-dollar computing costs ain't paying themselves, and last time we checked, you can't buy GPUs without…… serious money. 

Caught in 4K: Google contractors are reportedly using competitor Claude to improve Gemini’s outputs. 

The awkward part? Anthropic's terms straight up say "don't use Claude to train your competing models." Google's like "it's just evaluation, not training.”

Internal docs showing Claude references in Google's eval platform? Nothing to see here, folks. Move along.

Right?

What it means: 

Every AI company's using everyone else's models for “evaluation” like that friend who "borrows" Netflix passwords. 

This is tech's worst-kept secret since iPhone leaks. Google trying to act like they're not studying Claude's answers is giving major "I just happened to sit next to the smart kid during the test" energy.

Yeah, it’s a gray area. But the WFH and push for contractors to help train models means that a lot of lines are getting crossed. 

Google CEO Sundar Pichai is (finally) saying the quiet part out loud: AI is Google’s top priority moving forward. 

Reports are dishing out details from Google’s recent internal strategy meetings, and it’s Gemini focus All. Dang. Day. 

Sundar just told Google to make Gemini their 16th product to hit 500M monthly users. Why? Because he's tired of people thinking "AI = ChatGPT" like it's 2022 all over again.

They're promising Gemini will be a universal AI assistant that works across ALL your devices. 

And unlike ChatGPT's $200/month premium tier, Google's reportedly trying to keep Gemini cheap. 

What it means: 

NGL shorties….Google straight up fumbled the AI bag harder than a butterfingers convention. 

Their Gemini GTM strategy of 2023 and early 2024 was disastrous. (To say it nicely.) 

They had the users, the data, the compute - everything except the ability to ship product. 

Their biggest mistake? 

Hiding the real power of Gemini. 

Non-technical business leaders are the ones now making big decisions on which LLMs they’ll be using, and they’re often evaluating models on the front-end, not developer platforms. 

While players like Microsoft, OpenAI, Anthropic, Meta and Mistral display their most capable and updated models on the front end, Google has historically tucked their most recent models in their developer platforms. 

We saw a change in that strategy in December 2024. We’ll see if Google learned from (what we think will be a) trillion dollar mistake. 

OpenAI just un-ghosted their robotics program after a 2021 breakup, according to reports. 

But they're not just investing in Figure AI and 1X - they're going full Tony Stark and now looking to potentially build their own humanoid robots.

The company has re-established its robotics research group after closing it in 2021

This hits different with NVIDIA also pushing hard into robotics. It's like every AI company suddenly remembered they can't take over the world without bodies.

The timing's wild too - right when everyone's focused on their models and drama, they drop this bombshell.

What it means: 

Real talk? The next AI war ain't gonna be fought in your DMs - it's gonna be fought in your living room. 

When the company that made ChatGPT starts building robot bodies, that's your cue to start hitting the gym. We're going from "AI might take your job" to "AI might take your job AND your spot on the couch.” 

We just hope it’ll fold our laundry in addition to our spreadsheets.

A new Cambridge University study just confirmed what we all suspected: AI's getting scary good at manipulation. 

They're calling it the "intention economy" - where AI predicts and steers your choices before you even make them.

These systems aren't just reading your digital footprint - they're straight up forecasting and selling your future decisions to companies. Think targeted ads but with AI that knows you better than your therapist.

The researchers are saying advertisers can now use AI to adjust their strategies in real-time based on your behavior. Your wallet never stood a chance.

What it means: 

Sheeeesh - we went from "AI might manipulate us" to "AI's already got us in a chokehold" real quick. 

Social media had us hooked like digital dealers, but this intention economy? That's like giving AI the cheat codes to your brain. Your impulse purchases are about to get real impulsive.

The AI godfather just looked at humanity's odds and said "make it spicier." Geoffrey Hinton doubled his human extinction estimate from 10% to 20%. 

(Gulp.) 

He compared humans to toddlers next to AI, and after seeing some ChatGPT conversations, he might be generous. 

His key point? There's literally zero examples in history of something less smart controlling something more smart.

What it means: 

Geoffrey Hinton isn’t some wild AI doomsdayer. 

He’s literally one of the smartest people in the world when it comes to AI. Like… the Mount Rushmore of AI type. 

So when the brainiac who helped birth modern AI starts doubling down on extinction odds, that's like your doctor doubling your risk assessment while backing away slowly. 

But that 20% figure? 

Scary. 

Microsoft and OpenAI just turned the AGI philosophical debate into a quarterly earnings target: $100B in profit.

That’s reportedly the new target of when we’ve achieved Artificial General Intelligence. 

For context, Microsoft has reportedly invested more than $13 billion in OpenAI and has a 49% stake in the AI juggernaut. But, when AGI is “achieved” then Microsoft won’t be able to use post-AGI OpenAI tech in its products and services. 

So they switched from letting OpenAI's board make that AGI call to straight up putting a price tag on it.

So when/if an AI system can generate $100 billion in profit, then that means we’ve achieved AGI? 

Well…. That may be the price to pay to call something AGI, apparently. Lolz. 

What it means: 

They really said "forget the Turing test, show me the money." 

Big brain business move from Microsoft. 

TBH, we’ve been tip-toeing the AGI line for months. And as models like OpenAI’s o1 Pro and/or o3 get more capabilities, it’ll be hard to argue that we HAVEN’T achieved AGI. 

This is a smart move from Microsoft, as leaving the AGI determination to OpenAI’s board was risky AF to begin with. 

Imagine Microsoft Copilot losing access to next level AI tech with the snap of a few fingers? That’s some Thanos kinda power.