The AI Arms Race Just Got a New Battlefield: Your Attention
While you were busy trying to figure out which AI chatbot to use, OpenAI just decided to monetize your conversations. This week brought a seismic shift in how AI companies plan to make money, alongside some genuinely impressive technical launches that might actually change how you work.
Let’s cut through the noise and focus on what actually matters from the latest AI developments.
OpenAI Wants to Sell You Stuff While You Chat
The biggest news this week is OpenAI launching a self-serve advertising platform for ChatGPT. Yes, you read that right. The company that convinced millions to pay $20 monthly for ChatGPT Plus now wants advertisers to reach you while you’re asking about Renaissance art or debugging Python code.
According to reports from TechCrunch, the platform allows businesses to create sponsored responses and contextual ads within ChatGPT conversations. OpenAI is positioning this as “helpful recommendations” rather than traditional advertising, but let’s call it what it is: a new revenue stream as investor pressure mounts.
The pricing structure hasn’t been fully disclosed yet, but early access partners are reportedly paying CPM rates starting at $50 for targeted placements. That’s significantly higher than Google Ads but lower than premium LinkedIn advertising.
What This Actually Means for Users
If you’re a free ChatGPT user, expect to see more promotional content mixed into responses. Paid subscribers on Plus ($20/month), Team ($25/user/month), or Enterprise (custom pricing) plans will likely see fewer ads, though OpenAI hasn’t explicitly guaranteed an ad-free experience for paying customers.
The concerning part? This fundamentally changes the trust dynamic. When an AI recommends a product or service, will you know if that recommendation was organic or paid for?
Google Goes All-In on Agentic AI with Gemini 3.5
Google countered OpenAI’s monetization news with actual product innovation. The company announced Gemini 3.5 and Gemini Omni, marking what they’re calling the “agentic era” of AI.
Gemini 3.5 focuses on advanced reasoning and multi-step task completion. Unlike previous models that simply responded to prompts, this version can supposedly plan, execute, and adjust strategies across multiple tools and platforms. Think less chatbot, more digital coworker.
Gemini Omni takes things further by processing and generating across text, images, audio, and video simultaneously. It’s Google’s answer to GPT-4V and Claude 3, but with tighter integration into Google Workspace.
The Practical Applications
Google AI Studio users with Pro ($20/month) and Ultra ($30/month) subscriptions now get significantly increased usage limits. Pro users jumped from 1,000 to 2,500 requests per day, while Ultra subscribers can now make 5,000 daily requests.
This matters because previous limits made these tools impractical for serious work. Now you can actually build applications or run extensive research without constantly hitting rate limits.
Agentic AI Moves from Buzzword to Reality
Multiple sources, including Forbes and The Verge, highlighted that agentic AI is finally moving beyond tech demos into production environments. But what does “agentic” actually mean without the marketing fluff?
Agentic AI refers to systems that can take actions on your behalf with minimal supervision. Instead of asking an AI to write an email, you’d tell it to “coordinate next week’s meeting with the design team,” and it would check calendars, send invitations, and follow up with non-responders.
Who’s Actually Shipping Agentic Products
Anthropic recently updated Claude to handle multi-step computer tasks, allowing it to control your mouse and keyboard to complete workflows. Pricing starts at $0.008 per thousand input tokens for Claude 3.5 Sonnet.
OpenAI quietly expanded ChatGPT’s ability to trigger actions in connected apps through its plugin ecosystem, though adoption remains limited compared to Google’s Workspace integration.
Smaller players like Adept are building AI specifically designed to navigate software interfaces, though their products remain in limited beta.
The Model Wars Get More Expensive
AI Weekly reports that model competition is intensifying with record-breaking startup funding. Multiple AI companies raised significant capital this week, though specific numbers weren’t disclosed in available sources.
What we do know: the cost of training frontier models continues to climb exponentially. Wired reported that leading models now cost between $100-500 million to train, creating a massive barrier to entry that favors well-funded incumbents.
This consolidation matters because it determines who controls the infrastructure layer of AI. Right now, that’s primarily OpenAI, Google, Anthropic, and Microsoft.
What Happened to the Open Source Alternative
Meta continues releasing open-source Llama models, but they’re consistently 6-12 months behind closed-source competitors in capabilities. Llama 3 is impressive for an open model, but it’s not replacing GPT-4 or Claude 3 for demanding applications.
The performance gap means most businesses still choose paid API access over self-hosting open models, despite the long-term cost and control benefits of open source.
Video AI Reshuffles After Sora Shutdown
The video AI landscape shifted significantly this week following OpenAI’s decision to shut down Sora’s preview program. While OpenAI hasn’t provided a clear explanation, industry observers note timing coincides with increased compute costs and the ChatGPT advertising pivot.
Runway appears to be the main beneficiary. Their Gen-3 Alpha model is now the go-to for professional video generation at $0.05 per second of video. That’s expensive—a 10-second clip costs $0.50—but quality currently justifies the premium for commercial projects.
Pika offers a more affordable alternative at $10/month for 700 credits (roughly 140 seconds of video). Quality isn’t quite Runway level, but it’s sufficient for social media content and rapid prototyping.
The Stuff That Didn’t Make Headlines But Should
Superhuman quietly launched AI-powered email triage that actually works. It costs $30/month, but early users report saving 30+ minutes daily on inbox management.
Weave Robotics announced an $8,000 home robot that uses advanced AI for household tasks. That’s still expensive, but it’s the first sub-$10K option that doesn’t look like a toy. Expect this space to heat up rapidly as costs drop.
Scientists unveiled the “SpudCell” in what BBC called a major biotechnology breakthrough. While not directly AI-related, the discovery leveraged machine learning for protein structure prediction, demonstrating AI’s growing impact beyond software.
What These Tools Still Can’t Do
Despite the hype, current AI tools have significant limitations that rarely make it into launch announcements.
Agentic AI still requires extensive setup and frequently breaks on edge cases. You can’t just tell it to “handle my email” and walk away. You’ll spend hours configuring permissions, setting guardrails, and handling exceptions.
Video generation remains prohibitively expensive for most use cases. At current pricing, a 60-second commercial costs $3+ in generation costs alone, before accounting for the dozens of iterations needed to get usable results.
ChatGPT’s new advertising model introduces unknown biases into recommendations. There’s currently no way to distinguish between organic suggestions and paid placements, creating a trust problem that will take years to resolve.
The Bottom Line
This week marked a clear inflection point where AI companies shifted from pure capability races to business model experimentation. OpenAI’s advertising platform signals that subscription revenue alone won’t sustain these massive compute costs.
Google’s agentic AI push represents the more interesting development for people actually trying to use these tools productively. Increased API limits and multi-step reasoning capabilities make AI assistants genuinely useful rather than just impressive demos.
The video AI reshuffle after Sora’s shutdown shows how quickly this landscape changes. Tools you rely on today might disappear tomorrow as companies struggle with unit economics.
For anyone building on these platforms: diversify your dependencies, keep costs in check, and remember that today’s cutting-edge model is next month’s commodity. The only constant is that everything will change faster than you expect.
Disclaimer: Tool pricing and features change frequently. Always verify current information on official websites. Results vary based on individual use case.
Want More?
Ready to find your next favorite AI tool? Browse our full reviews at UntappedAI AI Tools.
Sources
openai.com • techcrunch.com • aistudio.google.com • forbes.com • theverge.com • anthropic.com • wired.com • bbc.com
