Dr. Dennis Weekly AI Newsletter — December 10, 2025

Happy New Week, Everyone!

Ever feel like you're running a marathon, only to realize you've been sprinting on a treadmill the whole time? That's the feeling many leaders got this week as Meta finally admitted its $70 billion metaverse bet was a costly diversion. The real race, as the past seven days made crystal clear, is in building AI that doesn't just answer questions, but actually thinks.

The Trend We're Loving: The Rise of Reasoning Engines

This week, the AI conversation shifted from how well AI can generate content to how well it can reason through complex problems. We're seeing the emergence of Reasoning Engines—AI models designed not just for speed and efficiency, but for deep, multi-step problem-solving. This is the leap from a calculator that can do math to a mathematician that can derive a new proof.

Why it matters:
Reasoning is the gateway to true AI-powered automation and strategy. Instead of just summarizing a document, a reasoning engine can analyze its strategic implications. Instead of just generating code, it can debug a complex system. This week, Google unveiled Gemini 3 Deep Think, a model that uses "parallel reasoning" to explore multiple hypotheses at once, tackling logic problems that stump other AIs [1]. Similarly, the new open-source DeepSeek V3.2 model integrates thinking directly into its tool-use, a major step toward more autonomous agents [2].

Key Implications:

1. From Task Automation to Workflow Automation: Reasoning engines can orchestrate complex, multi-step workflows that previously required human oversight. Google's new Workspace Studio is a prime example, allowing anyone to build AI agents that manage entire processes, not just single tasks [3].

2. More Reliable and Trustworthy AI: By showing their work and reasoning through problems, these models are less of a "black box," providing more transparent and verifiable outputs.

3. A New Class of AI Applications: Expect a wave of AI tools that can perform sophisticated analysis, from scientific research and engineering to financial modeling and legal discovery.

Further Reading: • Google's announcement of Gemini 3 Deep Think [1] • Technical breakdown of the DeepSeek V3.2 model [2]

Deep Dive

1. Meta's $70 Billion Hangover: The Metaverse Dream Dies for AI's Reality

Mark Zuckerberg finally blinked. After burning over $70 billion on a virtual world few wanted to visit, Meta announced this week it is slashing its Reality Labs budget by up to 30% and pivoting hard toward AI [4]. The move was a stunning, if predictable, admission that the company's all-in bet on the metaverse was a strategic blunder. Investors cheered, sending Meta's stock up 4% as the company redirected its focus and resources to what it's now calling "Superintelligence Labs."

For years, Zuckerberg preached the gospel of an immersive digital future, pouring billions into Quest headsets and the clunky, unpopulated Horizon Worlds. But with consumer adoption stalled and the AI revolution accelerating, the metaverse now looks like a costly distraction. The irony is thick: the very AI technologies that made the metaverse feel obsolete—powerful LLMs, generative content, and autonomous agents—are now Meta's saving grace. "What was once the company's defining bet has now become a cautionary tale," one analyst noted.

Now, Meta is in a race to catch up. With its Llama 4 model reportedly underperforming, the company is aggressively hiring top AI talent and investing billions in the compute infrastructure needed to compete with OpenAI and Google. The new strategy prioritizes near-term, AI-driven products like Meta AI assistants and smart glasses over the long-horizon, speculative future of VR. It's a humbling moment for a CEO who once staked his company's identity on a virtual dream, only to wake up to the stark reality of the AI arms race.

2. The AI Quality Wars: OpenAI's "Code Red" vs. Google's "Deep Think"

The AI cold war just got hot. This week, reports surfaced that OpenAI CEO Sam Altman has declared a "code red" to urgently improve ChatGPT's quality as Google's new Gemini 3 model gains momentum [5]. Altman has reportedly paused lower-priority projects like ad integrations and shopping agents to focus the entire company on boosting ChatGPT's core reasoning and response speed. The move is a clear sign that the days of OpenAI's uncontested dominance are over.

Just as OpenAI was rallying its troops, Google launched its most direct counter-attack yet: Gemini 3 Deep Think. This new, premium version of Gemini is explicitly designed for complex reasoning, using what Google calls "advanced parallel reasoning" to solve difficult logic, math, and science problems. "This new mode delivers a meaningful improvement in reasoning capabilities," said Tulsee Doshi, a senior director at Gemini, signaling Google's intent to compete not just on features, but on raw intellectual power [1].

This clash of the titans is forcing a new level of maturity on the AI industry. The focus is shifting from flashy demos to reliable, high-quality performance. For businesses, this means more powerful and specialized tools are on the way. For the AI giants, it means the pressure is on. As one OpenAI insider reportedly said, "The race is on, and we're running it like we're in second place."

3. The Enterprise Gets Its Own AI Factory: AWS Launches Nova Forge

For years, building a custom, frontier-level AI model was a privilege reserved for the tech giants with billions to spend. This week, Amazon's AWS blew that door wide open. At its re:Invent conference, AWS announced Nova Forge, a new service that lets enterprise customers train their own custom versions of Amazon's powerful Nova models for a flat fee of $100,000 a year [6]. It's a game-changing move that effectively democratizes access to high-performance AI.

Nova Forge solves a critical problem for businesses: when they try to customize existing models with their own data, the models often "forget" their core reasoning capabilities. As AWS CEO Matt Garman explained, "It's a little bit like humans trying to learn a new language. When you start when you're really young, it's actually relatively easy to pick up, but when you try to learn a new language later in life, it's actually much, much harder." Nova Forge allows companies to bake their proprietary data into the model during the training process, creating a powerful, custom AI that retains its intelligence.

Alongside Nova Forge, AWS also unveiled a new family of Nova 2 models, including a highly capable multimodal model (Nova 2 Omni) and a cost-effective reasoning model (Nova 2 Lite). With early customers like Reddit, Sony, and Booking.com already on board, AWS is making a bold play to become the go-to platform for enterprise AI. The message is clear: you don't need to be a tech giant to have a world-class AI model. You just need an AWS account.

“Dr. Dennis’ Picks This Week”

Tools shaping the AI future (and your workflow):

🔥 Google Workspace Studio – Build no-code AI agents to automate your entire workflow across Gmail, Drive, and Chat. https://workspace.google.com/studio

🚀 Runway Gen-4.5 – Create stunning, cinematic-quality videos from a simple text prompt with the world's top-rated video model. https://runwayml.com

🧠 Gemini 3 Deep Think – Tackle your most complex logic and reasoning problems with Google's most powerful AI thinking engine. https://gemini.google.com

🔧 Amazon Nova Forge – Train your own custom, enterprise-grade AI model for a fraction of the cost. https://aws.amazon.com/bedrock

🤔 Perplexity Memory – Get smarter, more personalized answers from an AI search engine that remembers your conversations and interests. https://perplexity.ai

Takeaways

This week wasn't about incremental updates; it was about fundamental shifts. The AI industry is moving beyond generative novelties and into the era of industrial-strength reasoning. The key insight for leaders is that AI is no longer just a tool for efficiency, but a platform for strategy. The one action to take now? Build a simple AI agent in Google Workspace Studio to automate a repetitive task. See for yourself how easy it's become to put AI to work. Next week, we'll dive into the explosion of AI-powered hardware.

Forward this to colleagues who need to stay ahead of the AI revolution. Have insights to share? Reply – I read every message.

Subscribe | Archive | LinkedIn | Twitter

References [1] Tom's Guide. (2025, December 5). Google just launched Gemini 3 Deep Think — its most powerful AI model yet. https://www.tomsguide.com/ai/google-gemini/google-just-launched-gemini-3-deep-think-its-most-powerful-ai-model-yet

[2] Raschka, S. (2025, December 3). A Technical Tour of the DeepSeek Models from V3 to V3.2. Sebastian Raschka's Magazine. https://magazine.sebastianraschka.com/p/technical-deepseek

[3] Google Workspace Blog. (2025, December 4). Introducing Google Workspace Studio to automate everyday work with AI agents. https://workspace.google.com/blog/product-announcements/introducing-google-workspace-studio-agents-for-everyday-work

[4] The AI Citizen. (2025, December 7). Top AI & Tech News (Through December 7th). https://theaicitizen.com/p/top-ai-tech-news-through-december-7thFor the first time, AI didn't just answer questions faster. It thought differently. Multi-step problem solving, autonomous reasoning, error correction—these aren't features. They're the foundation of a new category of intelligence.