Remember two years ago when Google was the sleeping AI giant and OpenAI was the untouchable king?

That narrative is officially dead and now OpenAI is all-in on ‘Code Red.’ 

This past week was absolute chaos in AI world. 

Google quietly released a model that is making OpenAI scream "Code Red" and panic-launch updates. Amazon is selling "Novellas" (yes, seriously). Anthropic is eyeing a staggering IPO.

The entire AI hierarchy just got reshuffled the past week, apparently.

If you are still betting on the old guard to stay on top, you are losing money.

Here is what you missed while the industry turned upside down.

Google very quietly released a new version of Gemini 3 this past week that is actually more powerful than the Pro version released just two weeks ago, which was already benched as the best LLM.

Yes. You read that right.

Google has launched its new "Deep Think" mode for Gemini 3 which is now available to AI Ultra subscribers. This model is designed to tackle complex math, science, and logic problems using parallel reasoning to explore multiple hypotheses at once.

It builds on the Gemini 2.5 architecture but features marked improvements in key benchmarks.

We are talking about a 93.8 score on GPT QA Diamond and a 45.1 score on Arc AGI 2 with code execution.

Early users are calling the performance "absolutely bonkers" despite some initial outages.

If you have used ChatGPT 5 Pro, this is very similar. It is a slow model.

But when it works? It is the most powerful large language model available.

What it means: 

Google has officially turned the tables on the model race.

While OpenAI was the clear leader for years, Google might have finally taken the crown for raw reasoning power.

If you need the absolute smartest model for complex logic, you now have to look at Gemini.

According to reports, the New York Times and the Chicago Tribune have filed lawsuits against Perplexity.

This marks a major escalation in the ongoing battle over news content and AI training data.

The New York Times alleges that Perplexity used its articles without permission to train AI models and power their chatbots. This is a very similar lawsuit to the one they filed against OpenAI previously.

The Times claims that Perplexity’s AI products reproduced their articles verbatim.

Even worse?

They allege the AI falsely attributed fabricated information to the newspaper.

Perplexity has been accused of bypassing paywalls and scraping content to generate answers for users which directly undercuts the publishers' business models.

It seems unlikely this specific case will go to trial. But whatever happens here will likely set the precedent for the future of AI and copyright law.

What it means: 

If a settlement forces companies to pay for content, that cost trickles down to literally every AI company.

Conversely, if Perplexity wins without a slap on the wrist, the entire concept of copyright in the US comes into question.

This lawsuit will define the future of the internet.

AWS introduced the Nova 2 family featuring four new AI models designed to handle complex tasks across text, images, video, and speech at their re:Invent conference.

The lineup includes Nova 2 Lite for cost-efficient reasoning and Nova 2 Pro for complex jobs like coding.

They also revealed Nova 2 Sonic for speech-to-speech AI and Nova 2 Omni for multimodal reasoning.

But here is the interesting part.

AWS also rolled out NovaForge.

This is a new service that allows enterprise customers to create customized versions of Nova models called "Novellas" for $100,000 a year. NovaForge gives companies access to pre-trained, mid-trained, or post-trained models while enabling them to use their own proprietary data.

It is a novel approach.

Amazon is trying to offer an "in-between" solution that isn't quite open source but offers more control than standard API access.

What it means: 

Amazon is betting big that enterprises want to own their models rather than just rent them.

By offering "Novellas," they are giving companies a way to build proprietary assets without starting from scratch.

We will see if companies actually adopt this or stick to the major frontier models.

Consulting giant Accenture is teaming up with OpenAI to roll out ChatGPT Enterprise to tens of thousands of its consultants.

This marks one of the largest enterprise adoptions of GenAI to date.

The partnership is expected to accelerate how Accenture develops consulting operations and tech services for clients.

According to CEO Julie Sweet, this move will help "accelerate enterprise reinvention in business outcomes."

OpenAI confirmed that Accenture will now have the largest number of professionals certified in OpenAI technologies.

Think back to two years ago.

Many large financial and consulting firms wrote off large language models and said they would never touch ChatGPT.

Yuuuuup.

Now they are adopting it at a massive scale because they realized they have no choice. You either adopt these tools or you get leapfrogged by smaller competitors who do.

What it means: 

The era of hourly billing for consulting is coming to an end.

Clients are getting smart and refusing to pay for hours when AI can do some standard consulting work in minutes instead of “billable” days.

Consulting firms must shift to outcome-based billing to survive this transition and it looks like Accenture is posturing to take a lead on AI-driven consulting.

Reports from the Financial Times indicate that Anthropic is making early moves toward a potential public offering.

The company is reportedly seeking a new funding round that could value it at over $300 billion.

They have even hired a law firm this month to begin laying the legal groundwork for an IPO.

While Anthropic says there are no immediate plans, this is standard behavior for a company prepping for the public market. Industry speculation suggests they could list shares as early as 2026.

This would put them in a race against rival OpenAI for the first spot on the public markets.

The company's valuation has soared from $183 billion in September to over $300 billion now.

What it means: 

Anthropic might beat OpenAI to the public market because their unit economics look better on paper. (For now.)

They are more profitable per user and have a clear path to breaking even by 2028.

OpenAI is burning cash faster and might need more time before they are ready for Wall Street.

Google has officially launched Workspace Studio to let employees build and manage AI agents to automate routine tasks.

This is a no-code automation builder similar to Zapier or Make but built directly into the Google ecosystem.

The system uses Gemini's multimodal reasoning to allow users with zero coding experience to create custom automations in minutes.

You can design automations just by describing your goals in natural language.

These agents integrate directly into Gmail, Drive, Chat, and other workspace apps.

What it means: 

The barrier to entry for building enterprise software continues to evaporate.

Everyday employees can now build their own automation tools inside of Google without waiting for IT.

This will lead to a massive explosion of productivity as non-technical staff start automating their own workflows.

Speaking of vibing and Google Replit and Google Cloud have announced a major expansion of their strategic partnership to bring advanced AI vibe coding tools to enterprise developers.

They pretty shook hands on bringing "vibe coding" to the masses.

Under this agreement, Replit will rely even more on Google Cloud’s infrastructure including Cloud Run and BigQuery.

They will integrate Google’s latest models like Gemini 3, Gemini 2.5 Flash, and Imagine 4 directly into the Replit environment.

Google 3 was recently added to Replit’s design mode to offer advanced capabilities for code generation.

Google has launched multiple vibe coding tools recently including Anti-Gravity, Opal, and Jules. This partnership confirms that Google sees this shift as permanent.

Everyday people are going to be building their own software using natural language and AI assistance.

What it means: 

Vibe coding is not a trend. It is the future of software development.

Google is betting its cloud infrastructure that natural language will become the primary programming language.

If you thought you needed to learn Python to build an app, you are already outdated.

The Verge reports that OpenAI might release GPT-5.2 as soon as tomorrow after has called an internal "code red" meeting due to the rapid rise of Google's Gemini models.

The alert reportedly called for an immediate shift in priorities across the company.

Teams have been directed to focus on improving speed, reliability, and reasoning power rather than developing new consumer features.

Staff were instructed to accelerate the deployment of the next major model update reportedly called GPT-5.2.

The target release date?

It could be as soon as tomorrow.

This update is designed to close the performance gap with Gemini 3. Internal sources say OpenAI's evals show GPT-5.2 could match or exceed Gemini 3's capabilities.

It is a reversal of roles from 2022 when Google called a code red after ChatGPT launched. Now Sam Altman is the one worried about losing the model race.

What it means: 

The features war is over. The reasoning war has begun.

OpenAI realized that flashy features like voice mode do not matter if your model isn't the smartest.

Expect a rapid series of model releases as these two giants try to one-up each other before the year ends.