To sum up the last week in AI news, you don’t have to look further than the great Ron Burgundy of Anchorman fame….. 

Well. That escalated quickly. 

The tech giants just turned AI development into a WWE Royal Rumble. 

Every. Single. Company. Just. SNAPPED. 

(Aside from Anthropic, which has been shipping all November.) 

OpenAI's doing "12 Days of Ship-mas," Google dumped their entire R&D vault into production, Meta's practically giving away state-of-the-art AI, and Amazon finally remembered they make more than just cardboard boxes. 

That’s just the teaser. 

Grab your popcorn - this is gonna be WILD.

Here’s what ya need to know.

OpenAI kicked off its 12 Days of OpenAI, or as many are calling it, the 12 days of Shipmas, with a bang last week. 

They announced both the full o1 model and a new ChatGPT Pro plan and dropped Sora access today. 

The O1 Pro model is rocking a 50% performance boost in reasoning capabilities compared to previous versions, using chain-of-thought processing that actually thinks before it speaks. 

This beast comes wrapped in a $200/month ChatGPT Pro subscription package that includes unlimited use, priority access, and computing power that'll make your local setup look like a calculator from 1995.

OpenAI announced their reinforcement learning fine-tuning research program. The O1 family's got three tiers: O1 Mini for the casual AI enthusiasts, full O1 for the serious players, and O1 Pro for the enterprises who treat token limits like suggestions.

Sam Altman's crew is specifically targeting complex problems in math, science, and coding with this update. Early testing shows it's crushing benchmarks harder than a caffeinated developer during launch week.

What it means: 

OpenAI's not just creating a digital caste system – they're building the whole socioeconomic ladder. Their tech is about to power everything from Microsoft's enterprise suite to Apple's new intelligence features, and they're pricing it like they know we can't live without it. 

The $20/month tier is about to feel like using dial-up in a 5G world.

Considering if you wanna get the most out of Sora and o1, you’ll need to spring for the pricey package. 

We covered it more in depth here. 

Microsoft just started rolling out Copilot Vision to Copilot Pro subscribers.  

Copilot Vision isn't just screen reading – it's understanding the full context of what you're doing online and interacting like a genius friend who actually read the manual. Currently rolling out to pro subscribers through Copilot Labs in the U.S., with a carefully curated list of supported websites.

The system can scan (select/supported) web pages, help plan activities, assist with shopping decisions, and actually understand the relationships between different elements on your screen.

Privacy features are locked down tight – all shared data gets deleted post-session, only Copilot's responses get logged for safety systems, and they're swearing on their cloud credits they won't use publisher data for training.

What it means:

Voice AI is about to make typing look like using an abacus. Microsoft's showing they've got their own AI genius bar, and they're not just riding OpenAI's coattails anymore. 

This is the future of work taking shape – imagine talking to an AI that actually understands your screen better than your remote coworker.

Google sure had a busy week in AI. Lolz. 

The new Gemini experimental 1206 model just leapfrogged OpenAI's ChatGPT 4 November version, continuing their back-and-forth benchmark battle that started with the 11/20 and 11/21 releases.

But here's the catch – you can only access this powerhouse in their developer sandbox. The front-end Gemini you're using is running on 3-9 month old models.

Their latest Android feature drop is supercharging Pixel phones with Gemini-powered abilities for calling contacts, drafting messages, and controlling device settings. 

DeepMind's Gencast is out here predicting weather patterns better than traditional models, though they're basing that claim on 2019 data published in Nature.

Veo, their new generative AI video model, is now in private preview on Vertex AI, offering businesses HD video creation from text or image prompts. 

Meanwhile, Genie 2 is a whole world simulator generating 3D environments from single images, complete with interactive objects and realistic physics.

Whew.

What it means: 

Google's shipping all this right after Sundar Pichai claimed AI development is slowing because "low hanging fruit has been picked." That's like saying you're taking a break while running a marathon.

These releases prove they're worried about losing their tech crown, no matter what Pichai says about slowing development. They're fighting a multi-front AI war and actually winning some of the races.

Elon's XAI just secured another $6B funding round, pushing their valuation past $40B.

This follows their previous $6B round from May that valued them at $24B, meaning they've nearly doubled their value in six months. 

The regulatory filing keeps most investor details under wraps, but Bloomberg's reporting suggests this round had way more small-time players than usual for deals this size.

Their model Grok is already showing promise, and with Elon's new government efficiency role, XAI's positioning itself as a serious contender in the AI race. 

What it means:

With this war chest, XAI's either about to revolutionize AI or burn through cash faster than a crypto startup in 2021. The small investor angle is clever – it's building a community of supporters while raising serious capital. 

Plus, with Elon's government connections growing stronger, XAI might have an easier time navigating future regulations.

Amazon's Nova family drops with four models: micro, lite, pro, and premier. 

These aren't just ChatGPT clones – they're rocking context windows that make OpenAI look antiquated. We're talking 128K tokens for micro and 300K for lite and pro, with plans to hit 2M tokens by early 2025.

Outta nowhere, Amazon is now a Tier 1 Frontier LLM player. 

Their Nova suite includes Nova Canvas for image generation and Nova Reel for 6-second videos, both packed with watermarking and content moderation features. They're being cagey about training data sources, hiding behind "competitive advantages" and legal concerns.

AWS is also promising a speech-to-speech model in Q1 2025 and an any-to-any model by mid-2025.

While they might be late to the party, they're making up for it by immediately competing at the highest level, potentially reducing their reliance on their $10B Anthropic investment.

What it means: 

Amazon watched everyone else's AI puberty and learned from their awkward phases. They're entering the market with features others are still prototyping, and their AWS integration means they've got a built-in enterprise customer base. 

This could be the beginning of Amazon reducing their Anthropic dependency, just like Microsoft's trying to wean off OpenAI a tad. 

ElevenLabs' new conversational AI platform speaks 31 languages and handles interruptions better than your most patient friend. 

It's got native Twilio support for texting and calls, plus both server-side and client-side tool calling capabilities.

The system integrates with all major LLMs – Claude, GPT, and Gemini – or you can plug in your own custom model. Their real-time model predicts when speakers are finished talking, making conversations feel natural instead of like you're talking to a robot having a stroke.

This isn't just another chatbot – it's a full-service platform that can handle customer support, outbound sales, scheduling, interactive gaming characters, and tutoring.

 Plus, you can upload your own knowledge base and set custom guardrails.

What it means: 

2025's customer service revolution is starting now. 

This is what happens when voice AI grows up and gets a real job. The ability to interrupt and have natural conversations means we're finally moving past the "please listen carefully as our menu options have changed" enjoyment.

David Sacks, former PayPal COO and All-In podcast cohost, is stepping in as the country’s AI czar alongside Elon Musk's new role leading government efficiency. 

Both are known OpenAI critics, with both specifically calling out their nonprofit-to-profit transition.

Sacks' VC firm Craft Ventures has invested in XAI, adding another layer to this spicy situation. While he's advocated for unregulated AI development markets, Musk has pushed for stronger AGI oversight.

This puts two influential tech figures who've been openly critical of OpenAI in positions to shape future AI policies, potentially affecting current regulations including Biden's AI executive order.

What it means: 

Having an AI czar who's more tech investor than AI researcher is like hiring a food critic to run a global restaurant group empire. 

Might not work out. 

The real story here is the potential conflict between OpenAI and these new government positions – especially since OpenAI's tech is becoming the backbone of major tech platforms.

OpenAI's hustling to find common ground with Microsoft on that pesky AGI clause that would block Microsoft from accessing their most advanced models post-AGI. 

In short, OpenAI is reportedly considering removing a clause in its Microsoft contract that would block Microsoft from accessing OpenAI's technology once artificial general intelligence (AGI) is achieved.

(And right now, OpenAI’s board is the one that gets to make that declaration. Lol) 

So, the existing agreement gives ownership of AGI to OpenAI's nonprofit board and prevents commercial use of the technology.

What it means: 

This is HUGE. 

This potential change is driven by OpenAI's growing need for capital, with CEO Sam Altman acknowledging they hadn't anticipated the massive funding requirements when they started. 

Microsoft, having already invested $13 billion in OpenAI, would benefit with this updated clause maintaining access to all future OpenAI technology, including AGI.

Meta just silently dropped Llama 3.3, and it's doing more with less like a Silicon Valley minimalist. 

Their 70B parameter model is matching the performance of the 405B parameter Llama 3.1, while costing just one cent per million tokens. 

That’s nutty. 

Also, according to Meta, the model achieved net zero emissions during training through renewable energy use, and it's available under their community license for most users.

What it means: 

Meta's showing that bigger isn't always better in AI – sometimes it's about how you use your parameters. They're making AI more accessible while the competition builds walled gardens, and they're doing it with a smaller environmental footprint. 

This is open(ish) source AI's glow-up moment.

At the end of today’s show, we dropped our predictions for the remaining 10 days, which is now 9. 

We predicted a Sora drop, which happened today. 

What’s next? 

We think the new Advanced Voice Mode with file uploads or maybe live vision that'll make Siri look like a speak & spell, and maybe that mysterious ‘Operator’ agent. 

Plus, we might see API costs drop again and possibly a GPT-4.5 update.

What do you think? 

Hit us with a reply and let us know.