Google’s Winning Week, OpenAI Skip Safety, and Kickboxing Robots: The AI Argument EP52

Google’s Gemini 2.5 isn’t just better, it might be in a league of its own. From coding to content creation, it’s outperforming everything else. And for once, nobody’s laughing at Google's AI efforts. While Justin’s all-in on the power and promise of Google’s new Agent framework, Frank’s still reeling from Google charging him €25 a pop to test VEO 2, and not even bothering with a warning label.

Overall, Google’s finally making good on its AI potential, rolling out powerful models, free dev tools, and smart protocols. Justin’s excited. Frank’s suspicious.

For developers and small teams, it’s a good time to explore. Just watch your wallet and don’t get too attached. Google have a history of spinning up projects and then killing them when we grow to love them.

Google’s not the only one in the spotlight either…

There’s a new approach to beating hallucinations by getting four LLMs to argue with each other before telling you anything. Meanwhile, OpenAI’s under fire for rushing safety checks, ChatGPT’s long-term memory has Frank twitching, and Meta’s boasting context windows big enough to fit your whole life story.

This one’s for founders, marketers, and anyone trying to work out where to place their bets as the AI race hits another gear.

00:34 So… was anyone not releasing AI?
10:55 Is Google's VEO 2 worth 25 bucks a prompt?
15:08 Can Tom & Jerry help AI make movies?
19:17 Can four LLMs make AI tell the truth?
22:40 Should we fear ChatGPT's long-term memory?
28:12 Is OpenAI skipping the safety checks?
31:15 Did Germany just prove UBI works?
33:48 Can we trust backdoored battle bots?

► SUBSCRIBE
Don't forget to subscribe to our channel stay updated on all things marketing and AI.

► STRATEGIC AND CREATIVE AI FOR SMALL BIZ MARKETING
For my full insights, be sure and subscribe to my emails here:
https://www.frankandmarci.com/ai/

► LINKS TO CONTENT WE DISCUSSED

Nvidia’s new Llama-3.1 Nemotron Ultra outperforms DeepSeek R1 at half the size
https://venturebeat.com/ai/nvidias-ne...

Deep Cogito emerges from stealth with hybrid AI ‘reasoning’ modelshttps://techcrunch.com/2025/04/08/dee...

Anthropic: Introducing the Max Plan
https://www.anthropic.com/news/max-plan

Kairos: Turn Repetitive Tasks into Automated Workflows.
https://www.kairos.computer/

Meta releases two Llama 4 AI models
https://www.theverge.com/news/644171/...

DeepCoder: A Fully Open-Source 14B Coder at O3-mini Level
https://www.together.ai/blog/deepcoder

Google Cloud Next 25
https://blog.google/products/google-c...

One-Minute Video Generation with Test-Time Training
https://test-time-training.github.io/...

Former Snap Exec Launches Higgsfield To Bring Cinematic Camera Language To AI-Generated Video
https://www.forbes.com/sites/charlief...

Hallucination mitigation using agentic ai natural language-based frameworks
https://arxiv.org/pdf/2501.13946

OpenAI slashes AI model safety testing time
https://www.ft.com/content/8253b66e-a...

Learn more about managing your memories in ChatGPT
https://help.openai.com/en/articles/8...

A German experiment gave people a basic monthly income. The effect on their work ethic was surprising
https://edition.cnn.com/2025/04/11/he...

China's Unitree robot is now training to fight in a ring.
https://x.com/minchoi/status/19105284...

Chinese robotics manufacturer left backdoor in product
https://www.axios.com/2025/04/01/thre...

► CONNECT WITH US
For more in-depth discussions, connect Justin and Frank on LinkedIn.
Justin:   / justincollery  
Frank:   / frankprendergast  

► YOUR INPUT
Would you pay €25 for a single prompt, even if the output was flawless? Or is that a deal-breaker?